4 10 46 https://eprints.ibu.edu.ba/files/original/cfd4bd924f50f430b9b77d3de65bb630.pdf 5df4ce187d54a38721f94c977d44e30a PDF Text Text Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 Information Systems Project Success Factors: Literature Review Nermina Durmic International Burch University, Sarajevo, Bosnia and Herzegovina nermina.durmic@ibu.edu.ba Abstract – The purpose of this paper is to identify and collect most commonly discussed project success factors in the context of information systems (IS) projects. Through the process of review of 88 books, relevant studies and scientific works 72 success factors were detected, with a total of 689 appearances, which are then classified into six factor groups: Planning, Project team, Project management, Development, Customer, Project facilitation. The paper reveals that factors that were recognized as the most critical ones for the success of information systems projects by majority of authors belong first to Planning, and then to Project team and Project management groups of factors. Findings in this paper are expected to serve as a valuable theoretical basis for future empirical research of success and failure of projects in modern information technologies (IT) organizations, and development of related IS project success models. Keywords – Project success, project success factors, information systems, literature review. 1. Introduction Despite efforts IT organizations are making today to survive and take a lead in the high competitive market, reports show that project success rates haven’t changed significantly over the past 15 years. Back in 2003, King [1] reported that in one IT organization three out of ten projects fail on average. In the same year, Lewis [2] reported that around 70% of all IS projects fail to fulfill the objectives set, where all failed and defectively completed projects were included. According to Arcidiacono [3], International Data Corporation published in 2009 that 25% of observed projects failed completely, and 50% of projects required rework. Eight years later, in 2017, PMI [4] performs a study observing “underperforming” organizations with less than 60% of project completed successfully, meeting their fundamental goals and business purpose. It is reported that 24% of projects in these organizations were completed within set timeframes, 25% within budget, 33% met original goals/business intent, 68% experienced scope creep, 24% were a complete failure and 46% of the budget was lost in case of project failures [4]. Considering the project definition adopted from Pinto and Slevin [5], explaining a project as an organization of people committed to common goals, involving valuable, high risk tasks of different sizes that have to be completed by previously set deadlines for a certain budget with high quality, and must have a very well defined objectives and sufficient resources necessary for task execution undertakings, it becomes clear that project success depends on many different factors. It’s very important to recognize and understand these factors because they serve as a guidance for definition of project management processes [6]. While investigation of �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 project success factors has been a topic of interest for many researchers and authors, there is still a lack of literature that summarizes findings in this context. This paper is intended to fill in this gap by collecting success factors in a broad range of relevant material in order to determine which aspects of project development process should be given the biggest attention to ensure the project success. After this introduction, a short background of the topic is presented, followed by explanation of the literature review methodology. Then, outcomes of the literature review process are presented in several segments, with short conclusion that gives suggestions for future research directions. 2. Background A. IS projects Most of the authors who contributed to the theory of project management by formulating a concrete definition of a „project“ agree that the definition is always based around five aspects, regardless of the field the project belongs to. Those are: people, project goals and requirements, project tasks, resources and various conditions. More specifically, a „project“ is generally defined as a complex organizational system of coordinated activities being performed in predefined order to achieve desired outcomes, in accordance with time and resource constraints ([7], [8], [9], [10], [11]). Information System (IS) projects on the other hand, which are objects of examination in this review, have more concrete characteristics. IS projects can be described as IT projects designed to answer the information processing needs of a certain organization. Attributes that make them different from any other non-IS project are three-fold: (1) they depend heavily on human resources and significant capital that is usually a constraint; (2) they are people oriented projects and their stakeholder teams are composed of three groups: development team members, managers, end users; (3) they are conceptual, meaning that IS projects can often be subjects to risks that come from stakeholder teams, their lack of knowledge or project type [12]. B. Project success criteria When it comes to defining the project success, the majority of researchers agree about the general success criteria: a successful project is a project completed on time, within scope and budget constraints ([13], [14], [15]). However, the application of this definition in real project environments is usually not that simple. The empirical research, conducted by Hussein [16], reveals that inadequate definition of the project success criteria is commonly a result of incomplete understanding of the project itself and setting unrealistic expectations about the benefits provided by project outcomes. Better understanding of project stakeholders' inputs, who have the impact on project context and who define the final outcome expectations, is recognized as the right approach to solving this challenge [16]. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 According to Frerer et al. [17] review of project success criteria literature, authors like Pinto and Slevin [18], Freeman and Beale [19], Khosravi and Afshari [20], Bryde and Robinson [21] define even five to nine criteria for measuring the project success. Collins and Baccarini [22] and Munns and Bjeirmi [7] find that there is a direct effect of project success towards project management success. Baccarini [13] concludes that there is no unique definition of project success that could be applied on any project, and that criteria specific for a given project needs to be defined at its inception point to ensure that all team members and stakeholders work in the same direction. 3. Methodology In order to respond to the research goal, it was necessary to review larger amount of literature, thus semisystematic review approach was selected as the most suitable. It helps discover theoretical aspects and teams, or common issues within a certain research discipline [23]. Books and journals suitable for review were collected from journal databases and online libraries. To shorten the material collection process search keywords like “information systems project success”, “project success factors”, “project success criteria” were used. 88 books, scientific works and articles were collected. Materials with repeating content adopted from previous literature were not taken in consideration. The review process was executed through three steps: (1) in the first literature walkthrough list of unique project success factors was created, 72 of them in total; (2) frequency of appearance of every factor was registered; (3) factors were grouped into 6 groups for easier interpretation: Planning, Project team, Project management, Development, Customer, Project facilitation. 4. Literature Review A. Project success seminal works According to Pinto and Prescott [24] critical success factors represent factors that lead to significant improvement of project implementation chances, if appropriately addressed. Therefore, many researchers have tried to recognize critical success factors that can be significant for all IS project in general. Leidecker and Bruno [25] explain these factors as variables or conditions that, if properly maintained, managed or sustained, can have a crucial impact on the organization success or failure. Definition of critical success factors may also depend on development of country, type of organization and business [26]. An overview of project success factors that have been discussed by the authors in previous studies on this topic is presented in the next sections. First studies in this field have started in early seventies, and in the next 20 years seven seminal works related to this topic have been created, written by seven different groups of authors. All other works that followed have considered these seven works as a starting point of their research. Seven seminal works and project success factors they define as critical ones for project success are listed below: �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 - Sayles and Chandler [27]: Capability and knowledge of a project manager, continuing involvement in execution of project activities, control subsystems, scheduling and time frames, monitoring activities and feedback provision. - Martin [28]: Project and organizational philosophy, project planning, definition of project goals, control and monitoring mechanisms, top management support, authority delegation and organization, project team, resource allocation. - Cleland and King [29]: Project schedule, project scope, process of task execution, financial support, logistic requirements, facility support, market intelligence (who is the client), executive development and training, manpower and organization, acquisition, information and communication channels, project review. - Baker et al. [30]: Project manager, skills and knowledge of the project team, team commitment, project goals, project planning, budget and cost estimates, monitoring and control techniques, project inception difficulties, inadequate hierarchy. - Lock [31]: Team commitment to project activities, top management support, communication procedures, progress review meetings, control mechanisms, project manager - Morris and Hough [32]: Technical complexity, project goals, project scheduling, budgeting, legal issues, implementation issues, project idea and innovation. - Pinto and Slevin [33]: Client involvement, human resources, monitoring and control, project team lead, communication, issue handling, technical difficulties, urgency, project politics, environmental impact. B. Project success factors In this section 72 project success factors collected through literature review process and their groupings are presented. Graph on Figure 1 shows the comparison of number of different success factors contained in every one of the six factor groups, while graph on Figure 2 shows the frequency of appearance of success factors in each group in the reviewed literature. Figure 1. Number of success factors in each group of factors �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 Figure 2. Frequency of appearance of success factors in literature Charette [34] claims that IS projects almost never fail for only one or two reasons. According to him the reason of projects failing is a combination of few factors from the following list: bad estimation of necessary resources, poorly defined project requirements, unrealistic expectations and project goals, absence of communication between customers, poor management of users and developers, inadequate technology, lack of a good reporting of the project status, poor risk management, low quality of project management and development practices, project complexity. Ewusi-Mensah [12] agrees with him, indicating that the cancelation of IS development projects is usually caused by few combined factors, such as: project objectives, knowledge and skills of a project team, monitoring and control, lack of involvement of the top management, project costs and deadlines. Both authors consider definition of project goals and management as the most significant factors. Moohebat et al. [26] did an investigation about the country difference impact on project success factors, considering developed and developing countries. According to their research, the only project success factor, which is of equal importance for both groups of countries, is the top management support. Procaccino et al. [35] focus on the project success factors strictly from developers' point of view. They find that successful project for a developer means a project that is managed the way that ensures that development team has enough of necessary resources and the least possible amount of distractions when executing their daily jobs. For them, involvement of the customer in the project execution who is available to give feedback for the work done, and well defined project scope lead to successful project outcomes [35]. Egorova et al. [36] focus on stakeholders' point of view, dividing them into three groups: strategic-view stakeholders, operational-view stakeholders and tactic stakeholders. In their work, Egorova et al. [36] state that both operational and strategic stakeholders place „understanding the customer's problems“ to the first place. Operational respondents give a special attention to good programming, and strategic respondents see „customer involvement“ and „completed and accurate requirements“ as more important factors. Tactic stakeholders choose „very good project management“ as the most important factor for the project success. For both operational and tactic respondents, „team experience“ plays the essential role. Frese and Sauter [37] divided reviewed projects into failed, challenged and successful projects groups, aiming to see if there are common factors affecting project outcomes in all three groups. The conclusion they draw is that the quality of customer involvement and requirements definition affect the project's final status in all three groups. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 Planning factors In 95% of reviewed literature the Planning group of factors is discussed as the essential one for the project success. Project planning, scheduling and control, project requirements and scope, project goal, mission and vision are recognized as leading success factors in this group by majority of authors. All Planning success factors are listed in Table 1. Table 1. Planning group of project success factors Planning success factors Project planning, scheduling and control Requirement specification and scope Definition and understanding of project goals, mission and vision Budgeting – cost estimates Project/technical complexity Process and working procedures Time estimations Project organizational philosophy/ organization structure Realistic expectations Project itself/ project idea Project strategic fit Project size Project pace TOTAL (max = 88) Source [5], [18], [27], [28], [29], [30], [31], [32], [33], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74] [11], [12], [18], [20], [26], [34], [35], [36], [37], [38], [39], [40], [43], [45], [47], [49], [50], [52], [53], [55], [56], [58], [60], [62], [67], [68], [69], [70], [71], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83], [84] [5], [11], [12], [18], [26], [28], [30], [32], [33], [34], [37], [43], [45], [47], [48], [50], [52], [58], [60], [61], [63], [66], [67], [68], [69], [72], [74], [75], [76], [77], [81], [83], [85], [86], [87], [88], [89], [90], [91] [12], [29], [30], [36], [39], [45], [50], [51], [53], [59], [60], [62], [64], [67], [68], [69], [71], [73], [74], [78], [81], [89] [12], [34], [47], [48], [63], [69], [70], [74], [78], [79], [81], [82], [87], [90], [92] [6], [34], [35], [49], [64], [66], [67], [68], [69], [70], [74], [75], [81], [87], [93] [12], [22], [35], [36], [50], [53], [71], [83], [94] [6], [12], [28], [29], [38], [57], [59], [73], [95] [36], [37], [58], [71], [75], [81], [96], [97] [6], [29], [49], [70], [76], [95] [6], [26], [38], [60], [75] [78], [87] [59], [92] Freq. 49 % 56% 41 47% 39 44% 22 25% 15 17% 15 17% 10 9 11% 10% 8 6 5 2 1 84 9% 7% 6% 2% 1% 95% Morisio et al. [45] write that definition of requirements before projects starts or, if not possible, their completion in the initial phases is a factor of success, which supports the “Glass law” which says that insufficiently defined requirements are the major reason for project failures. Zouaghi and Laghouag [11] find that clear definition of needs through requirements is one of three factors that present a high risk for the final result of a project. According to Kappelman et al. [52], not documenting the functional performance and reliability of requirements and scope is an early warning sign of IS project failure, which shouldn't be ignored. Definition of requirements is also stated by Frese and Sauter [37] as a common factor for successful, challenged and failed projects. Nasir and Sahibuddin [74] rated the clear requirements and specifications factors as the most important ones among all project success factors. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 Reel [75] sees project complexity as the basic problem of computing in the context of project development. Handerson [78] agrees with him, saying that complexity, together with the project size, is the main reason why large IT projects fail. On the other hand, Nguyen [87] does agree that complexity, from technical perspective, makes a strong negative effect on project success, but unlike Handerson [78], he rates project size as a factor that almost doesn't affect the project success. Ogden [76] finds that the project idea is a success factor, but not a very important one for the project success. Hirshfield and Lee [96] say that successful projects are ones with realistic expectations and timeframes, and suggests „planning in advance“ as an activity that should ensure meeting schedule related conditions [96]. Project team factors As presented in Table 2, 89% of reviewed literature detected Project Team related factors as the most common success factors in an IS project creation. Role of a project manager, team commitment and communication are recognized as leading success factors in this group. Morisio et al. [45] asserts that human factors play a key role in software development. Zouaghi and Laghouag [11] and Ogden [76] also put the accent on productivity and motivation of the project team and their crossfunctionality. According to the research study of Wong et al. [54], poor project manager's effectiveness serves as a critical project failure factor. Manager's capability and skills are recognized by Nguyen [87] and Nasir and Sahibuddin [74] as a strong positive effect on a project's success. Perkins [98] states that the major cause of project failure is project manager not having the required knowledge, or not being able to apply it appropriately. While many researchers share opinion that project manager's field experience is also very important, Kaya et al. [70] disagree with that. Surprisingly, researchers in only 2% of reviewed literature found that working environment is a factor that affects the success of an IS project. Unlike Nguyen [87] and Pinto and Slevin [33], Nasir and Sabihuddin [74] even find environmental influences as completely unsubstantial factor. Hong et al. [73] suggest that it's important to have a good communication among all related parties including planners, consumers and developers for establishment of a good project model. Table 2. Project team group of project success factors Project team success factors Project manager Team commitment Source [12], [26], [29], [30], [31], [35], [36], [37], [38], [39], [42], [43], [47], [49], [50], [52], [54], [55], [57], [70], [71], [74], [76], [77], [78], [80], [83], [85], [87], [88], [90], [91], [92], [94], [98], [99] [6], [11], [26], [27], [28], [30], [32], [31], [33], [35], [37], [38], [39], [40], [41], [44], [48], [52], [55], [58], [59], [61], [66], [70], [71], [72], [74], [75], [76], [93], [100] Freq. 36 % 41% 31 35% �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 Communication Knowledge and skills of development team Capability, skills and experience of a project manager Team composition and assembly Personal Recruitment/ Ambition Education and training availability Team motivation and productivity Losing skilled team members Experience of development team External consultant Knowledge sharing Teamwork and collaboration Team building Personal interests Knowledge application Working environment Adding new team members Access to talented people Best practices and lessons learned TOTAL (max = 88) 31 35% 21 24% 21 24% 11 13% 5 6% 5 6% 4 5% [45], [52], [54], [75] 4 5% [12], [36], [47] 3 3% [49], [69], [76] [38], [54], [101] [41], [78] [12], [47] [44], [98] [98], [56] [33], [87] [45] [47] [75] 3 3 2 2 2 2 2 1 1 1 3% 3% 2% 2% 2% 2% 2% 1% 1% 1% 78 89% [5], [12], [26], [29], [31], [33], [34], [36], [37], [38], [39], [43], [47], [48], [52], [53], [56], [60], [61], [63], [66], [67], [69], [70], [72], [73], [74], [76], [79], [90], [93] [12], [30], [37], [41], [42], [49], [52], [53], [62], [66], [67], [68], [69], [70], [72], [74], [78], [81], [82], [87], [98] [27], [33], [36], [37], [45], [49], [50], [55], [61], [63], [66], [67], [68], [69], [70], [72], [74], [81], [82], [95], [98] [5], [12], [36], [38], [66], [69], [74], [82], [86], [90], [91] [12], [30], [41], [60], [77] [29], [47], [70], [88], [98] [11], [59], [74], [81] Project management factors Project management activities are defined as success factors in 85% of the reviewed literature (Table 3). In this group, top management support and effective monitoring and reporting are recognized as leading success factors. Whittaker [50] avers that inadequate risk management is among biggest IS project failure reasons. As the organization gets bigger, risk management becomes more significant factor of success. Taylor [51] states that inability to manage the risk and project related uncertainties has been frequently recognized as a critical segment of IS project management. Kappelman et al. [52] find the lack of support of top managers as an extremely important early warning sign of IS project failure, even the most important among other factors. Nguyen [87] emphasizes that good management in general is essential for a project to succeed, especially human resources management, quality management and time management. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 Table 3. Project management group of project success factors Project management success factors Top management support Monitoring and reporting Change management Project risk management Effective leadership Executive management support Quality management Time pressure Measurement systems Authority delegation Time management Late failure warning signals Success criteria Definition Overtime handling TOTAL (max = 88) Source Freq. % [5], [11], [12], [26], [28], [33], [34], [35], [36], [47], [48], [50], [52], [54], [55], [57], [60], [61], [66], [67], [68], [69], [70], [72], [73], [74], [78], [81], [91], [92], [94] [12], [26], [27], [28], [29], [33], [34], [47], [48], [50], [55], [56], [59], [60], [63], [64], [65], [67], [68], [69], [70], [71], [72], [74], [79], [81], [93], [98] [26], [37], [41], [42], [44], [46], [47], [50], [52], [55], [56], [61], [64], [67], [69], [71], [74], [75], [79], [89], [90], [91] [6], [11], [14], [34], [38], [41], [45], [47], [50], [51], [55], [59], [67], [74], [78], [80], [92], [101], [102] [12], [38], [41], [47], [63], [67], [69], [71], [74], [90], [99], [100], [103] [12], [37], [44], [58], [71], [83], [97], [100] 31 35% 28 32% [41], [62], [64], [65], [70], [74] [33], [34], [75], [76] [12], [31], [76] [28], [31] [39], [78] [53] [52] [45] 22 19 21% 13 15% 8 9% 6 4 3 2 2 1 1 1 75 7% 5% 3% 2% 2% 1% 1% 1% 85% Development factors Development related factors are recognized as ones critical for project success in 46% of the reviewed literature. Technology and tools, together with the availability of adequate resources are recognized as leading success factors in this group. Yet, many authors agree that success of technical development depends on the proper project planning phase. May [53] concludes that projects with inflexible technical architecture and undefined guidelines for managing the project technical requirements have high risk of failures. According to him the key of success lays in correct handling of technical aspects of the project. White and Fortune [47] underline that quality of planning must be taken into account peculiarly to have a successful development phase, and that it's extremely important that schedule of development activities is realistic. All detected Development group success factors are listed in Table 4. Table 4. Development group of project success factors Development success factors Source Technology, tools [12], [32], [34], [49], [50], [57], [59], [61], [67], [68], [69], [71], [74], [75], [81], [82], [88], [91], [92] [28], [34], [37], [47], [49], [53], [65], [66], [67], [71], [72], [74], [76], [81], [82], [94] Adequate resources availability Freq. % 19 22% 16 18% �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 Development approach IT infrastructure Testing, verification and validation Programming Data quality and integrity Architecture and design Technical tasks Interface with other projects Facility support TOTAL (max = 88) [18], [74], [83], [91] [54], [64], [74], [83] [26], [54], [79], [80] 4 4 4 5% 5% 5% [26], [36], [80] [26], [91] [53], [80] [30], [33], [60] [57] [29] 3 2 2 2 1 1 40 3% 2% 2% 2% 1% 1% 46% Customer factors Customer related group of factors is recognized as an important one for IS project success in 42% of reviewed literature, having the overall customer involvement as the most critical factor. Frese and Sauter [37] find that certain level of user involvement is a prevalent factor of project success and failure. According to Hirshfield and Lee [96] project team with their project manager can be sure their project meets its goals only if end users are involved in the process. On the contrary, Nasir and Sahibuddin [74] claim that project champion is not important in project development process at all. All detected success factors in the Customer group are presented in Table 5. Table 5. Customer group of project success factors Customer success factors Customer involvement Customer approval Involvement of project champion Inflexible customer TOTAL (max = 88) Source [5], [11], [33], [35], [36], [37], [42], [45], [47], [48], [49], [52], [54], [59], [60], [61], [66], [67], [68], [69], [71], [72], [73], [74], [82], [83], [84], [85], [88], [94], [96], [97], [100] [5], [33], [36], [72] [26], [49], [58] [54] Freq. 33 % 38% 4 3 4% 3% 1 37 1% 42% Project facilitation factors Project facilitation factors are recognized as project success factors in 17% of reviewed literature. Although the most significant literature about project management and project success doesn’t write about troubleshooting, conflict handling or external influences as project success factors a lot, the 10-factor model of the project development process defined by Pinto and Slevin [33] lists troubleshooting as its tenth component. Human error factor is recognized as the success factor by White and Fortune [47] and Levenson [79], but Attarzadeh and Ow [13] who also discuss human error factors as significant ones for the project success, claim that the reason for the human error is the cause of bad project management and inability of responsible roles to convert the theory of project management into practice. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 All detected success factors in the Project facilitation group are listed in Table 6. Table 6. Project facilitation group of projects success factors Project facilitation success factors Troubleshooting Conflict handling External influences Human error factor Tolerance of bad news Technical difficulties Start-up difficulties TOTAL (max = 88) Source [5], [26], [33], [48], [70], [72] [39], [47], [53] [47], [95], [98] [47], [79] [90] [89] [30] Freq. % 6 3 3 2 1 1 1 15 7% 3% 3% 2% 1% 1% 1% 17% 5. Conclusion The paper summarizes findings of the review of 88 works that discuss factors that affect the success of IS projects. While each of 72 detected success factors can play a major role for the outcome of an IS project, the majority of authors agree that project planning, scheduling and control; requirements and defined project scope; definition and understanding of project goals, mission and vision; role of a project manager; team commitment; communication and involvement of customers belong to the most significant ones. On the other hand, technical difficulties; overtime work; project size and strategy; development team dynamics; overtime handling; project architecture and design are the least frequent success factors in reviewed literature. Detected factors are classified into 6 groups, according to the segment of IS project development process these factors may have an impact on. The outcome shows that Project team group of factors is found to be the most diverse and highly significant, while Planning group of factors is recognized to be the most significant one for the IS project success. During the review process, it is noticed that not a lot of empirical research was conducted on this topic in the last 10 years, which results in the lack of literature with fresh findings and conclusions published in this period. This paper provides a foundation for conducting such empirical research as a future research direction, with two-fold goal: (1) to discover if recent trends in the process of IS project development – like Agile methodologies, and unpredictable dynamic of software market today, resulted in significant changes in the list of key project success factors and definition of “project success” overall; (2) to establish a project success model that can be used as a guidance in the process of IS project development. REFERENCES [1] J. King, Survey shows common IT woes persist [Online]. Available: http://www.computerworld.com/s/article/82404/Survey_shows_common_IT_woes_persist?taxonomyId=014 �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 [2] B. Lewis, The 70-percent failure [Online]. Available: http://archive.infoworld.com/articles/op/xml/01/10/29/011029/opsurvival.xml [3] G. Arcidiacono, „Comparative research about high failure rate of IT projects and opportunities to improve,“ PM World Journal, vol. 6, no. 2, February, 2017, pp. 1-10. [4] PMI „Success Rates Rise. Transforming the high cost of low performance,“ PMI's Pulse of the Profession. 9th Global Project Management Survey, Project Management Institute, 2017. [5] J. K. Pinto, D. P. Slevin, “Critical Factors in Successful Project Implementation,” IEEE Transactions on Engineering Management, vol. 34, no. 1, February, 1987, pp. 22-27. [6] X. Pu, X. Li, H. Zhu, A cross-national study of success factors in innovation project: China and western developed countries [Online]. Available: https://pdfs.semanticscholar.org/a0dc/23e82971036e84e49392495f3b0f2138f028.pdf [7] A. K. Munns, B. F. Bjeirmi, “The role of project management in achieving project success,” International Journal of Project Management, vol. 14, no. 2, April, 1996, pp. 81-87. [8] G. Lowery, Managing Projects With Microsoft Project 4.0: For Windows and MacIntosh, 1 st ed. New York: John Wiley & Sons, Inc., 1997. [9] C. Morley, Management d'un projet système d'information : Principes, techniques, mise en oeuvre et outils, 7th ed. Dunod, 2012. [10] A. Diallo, D. Thuillier, “The success of international development projects, trust and communication: an African perspective,” International Journal of Project Management, vol. 23, no. 3, April, 2005, pp. 237– 252. [11] I. Zouaghi, A. Laghouag, Empirical Study of Key Success Factors in IS Projects [Online]. Available: https://halshs.archives-ouvertes.fr/halshs-00662037/document [12] K. Ewusi-Mensah, „Critical Issues in Abandoned Information Systems Development Projects,“ Communications of the ACM, vol. 40, no. 9, September, 1997, pp. 74-80. [13] I. Attarzadeh, S. H. Ow, “Project Management Practices: The Criteria for Success or Failure,” Communications of the IBIMA, vol. 1, 2008, pp. 234-241. [14] K. Bakker, A. Boonstra, H. Wortmann, “Does risk management contribute to IT project success? A meta-analysis of empirical evidence,” International Journal of Project Management, vol. 28, no. 5, July, 2010, pp. 493-503. [15] J. A. Hoffer, J. F. George, J. S. Valacich, Modern Systems Analysis and Design, 7 th ed., Pearson Education Inc., 2013 [16] B. A. Hussein, (2013) “Factors Influencing Project Success Criteria,” in The 7 th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, Berlin, Germany, 2013. [17] A. A. Frerer, M. Mahmoud, H. Haleema, R. Aalmamlook, “Overview success criteria and critical success factors in project management,” Journal of Industrial Engineering and Management, vol. 7, no. 1, April, 2018, pp. 1-6. [18] J. K. Pinto, D. P. Slevin, “Project success: definitions and measurement techniques,” Project management journal 2, January, 1988, pp. 67-72. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 [19] M. A. Freeman, P. Beale, “Measuring project success,” Project management journal, vol. 23, no. 1, March, 1992, pp. 8-17. [20] S. Khosravi, H. Afshari, “A success measurement model for construction projects,” in International conference on financial management and economics, Singapore, Singapore, 2011, pp. 186-190. [21] D. J. Bryde, L. Robinson, “Client versus contractor perspectives on project success criteria,” International journal of project management, vol. 23, no. 8, November, 2005, pp. 622-629. [22] A. Collins, D. Baccarini, “Project success – a survey,” Journal of construction research, vol. 5, no. 2, September, 2004, pp. 211-231. [23] H. Snyder, “Literature review as a research methodology: An overview and guidelines”, Journal of Business Research, vol. 104, November, 2019, pp. 333 – 339. [24] J. Pinto, J. Prescott, “Variations in Critical Success Factors over the Stages in the Project Lifecycle,” Journal of Management, vol. 14, no. 1, March, 1988, pp. 5-18. [25] J. K. Leidecker, A. V. Bruno, “Identifying and using critical success factors,” Long Range Planning vol. 17, no. 1, February, 1984, pp. 23-32. [26] M. R. Moohebat, A. Asemi, M. D. Jazi, (2010) „A Comparative Study of Critical Success Factors (CSFs) in Implementation of ERP in Developed and Developing Countries,“ International Journal of Advancements in Computing Technology, vol. 2, no. 5, December, 2010, pp. 99-110. [27] L. R. Syles, M. K. Chandler, Managing large systems: organizations for the future, 1 st ed. New York: Harper & Row, 1971 [28] C. C. Martin, (1976) Project Management, Hilghlighting ed. New York: Amacom Books, 1976. [29] D. I. Cleland, W. R. King, System Analysis and Project Management, 3 rd ed. New York: McGraw- Hill, 1983. [30] B. N. Baker, D. C. Murphy, D. Fisher, “Factors Affecting Project Success,” in Project Management Handbook, D. I. Cleland, W. R. King, 2nd ed. Chichester, United Kingdom, 1997, ch. 35, pp. 902-919. [31] D. Lock, Project Management, 3rd ed. New York: St Martin’s press, 1984. [32] P. W. G. Morris, G. H. Hough, Preconditions of success and failure in major projects, Technical Paper/Major Projects Association, no. 3, Oxford: Templeton College, the Oxford Centre for Management Studies, 1987. [33] J. K. Pinto, D. P. Slevin, “Critical success factors in R&D projects,” Research Technology Management, vol. 32, no. 1, January, 1989, pp. 31-35. [34] R. N. Charette, “Why Software Fails,” IEEE Spectrum, vol. 42, no. 9, September, 2005, pp. 42-49. [35] J. D. Procaccino, J. M. Verner, S. P. Overmyer, “Case Study: Factors for Early Prediction of Software Development Success,” Information and Software Technology, vol. 44, no. 1, January, 2002, pp. 53-62. [36] E. Egorova, M. Torchiano, M. Morisio, C. Wohlin, A. Aurum, R. B. Svesson, “Stakeholders Perception of Success: an Empirical Investigation,” in 35th Euromicro Conference on Software Engineering and Advanced Applications, Patras, Greece, 2009. [37] R. Frese, V. Sauter, „Project success and failure: What is success, what is failure, and how can you improve your odds for success?“ [Online]. Available: http://www.umsl.edu/~sauterv/analysis/6840_f03_papers/frese/ [38] E. Camilleri, Project Success – Critical Factors and Behaviours, 1st ed. New York: Routledge, 2011. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 [39] J. Gido, J. P. Clements, Successful Project Management, 6th ed. Cengage Learning, 2014 [40] C. M. Padgett, The Project Success Method: A Proven Approach for Achieving Superior Project Performance in as Little as 5 Days, 1st ed. Wiley, 2009. [41] J. R. Turner, Handbook of Project Management, 4th ed. Hampshire, England: Gower Publishing Limited, 2007. [42] A. Hashim, B. Allan, Treasury Reference Model, World Bank Technical Paper, no. 505, Washington, DC: World Bank, 2001. [43] E. Verzuh, The Fast Forward MBA in Project Management, 4th ed. Hoboken, New Jersey: John Wiley & Sons, 2012. [44] H. Kerzner, Advanced Project Management: Best Practices on Implementation, 2nd ed. Wiley, 2008. [45] M. Morisio, E. Egorova, M. Torchiano, (2007) “Why software projects fail? Empirical evidence and relevant metrics,” Proceedings of the IWSM – Mensura, Palma de Mallorca, Spain, 2007, pp. 299 – 308 [46] D. Dvir, T. Lechler, “Plans are nothing, changing plans is everything: the impact of changes on project success,” Research Policy, vol. 33, no. 1, January, 2004, pp. 1-15. [47] D. White, J. Fortune, “Current practice in project management – an empirical study,” Internal Journal of Project Management, vol. 20, no. 1, January, 2002, pp. 1-11. [48] J. K. Pinto, D. P. Slevin, “Critical Success Factors in Effective Project Implementation,” in Project Management Handbook, D. I. Cleland, W. R. King 2 nd ed. John Willey & Sons, Inc., 2008. [49] C. H. Loch, H. J. Kayser, Do general project success drivers in product development exist? An Empirical Study of a European Technology Manufacturer, 1998. [50] B. Whittaker, “What went wrong? Unsuccessful information technology projects,” Information Management & Computer Security, vol. 7, no. 1, March, 1999, pp. 23-30. [51] H. Taylor, “Project Management and Problem Resolution Strategies for IT Projects: Prescription and Practice,” Project Management Institute, vol. 37, no. 5, December, 2006, pp. 49-63. [52] L. A. Kappelman, R. McKeeman, L. Zhang, “Early Warning Signs of IT Project Failure: The Dominant Dozen,” Information Systems Management, vol. 23, no. 4, Fall, 2006, pp. 31-36. [53] L. J. May, “Major Causes of Software Project Failures,” Crosstalk: The Journal of Defense Software Engineering, July, 1998, pp. 9-12. [54] A. Wong, H. Scarbroug, P. Y. K. Chau, R. Davison, “Critical Failure Factors in ERP Implementation” in Pacific Asia Conference on Information Systems, Bangkok, Thailand, 2005. [55] M. Boadle, „Project Failure and Success Factors“ [Online]. Available: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.199.5692&rep=rep1&type=pdf [56] J. Kanter, J. J. Walsh, “Toward more successful project management,” Information System Management, vol. 21, no. 2, 2004, pp. 16-21. [57] W. J. Mengesha, (2004) “Performances for Public Construction Projects in Developing Countries: Federal Road & Educational Building Projects in Ethiopia,” PhD. Dissertation, Norwegian University of Science and Technology, Trondheim, Norway, 2004 [58] inc.com/ J. Sturdivant, The CNSI Requirements Analysis Process [Online]. Available: https://www.cns- �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 [59] P. W. G. Morris, “Project management: a view from Oxford,” International Journal of Construction Management and Technology, vol. 1, 1986, pp. 36-52. [60] O. Zwikael, S. Globerson, “From critical success factors to critical success processes,” International Journal of Production Research, vol. 44, no. 17, September, 2006, pp. 3433 – 3449. [61] C. Dong, K. B. Chuah, L. Zhai, “A Study of Critical Success Factors of Information System Projects in China,” Proceedings of the PMI Research Conference, London, July, 2004. [62] B. W. Boehm, “Software Risk Management Principles and Practices,” IEEE Software, vol. 8, no. 1, January, 1991, pp. 32-41. [63] W. S. Humphrey, (2005) “Why Big Software Projects Fail: The 12 Key Questions,” CrossTalk: The Journal of Defense Software Engineering, vol. 18, no. 3, March, 2005, pp. 25-29. [64] C. Jones, “Patterns of Large Software Systems: Failure and Success,“ Computer, vol. 28, no. 3, March, 1995, pp. 86-87. [65] C. Jones, “Our Worst Current Development Practices,“ IEEE Software, vol. 13, no. 2, April 1996, pp. 102-104. [66] K. Milis, R. Mercken, “Success Factors Regarding the Implementations of ICT Investment Projects,“ International Journal of Production Economics, vol. 80, no. 1, November, 2002, pp. 105-117. [67] E. Oz, J. J. Sosik, (2000) “Why Information Systems Projects are Abandoned: A Leadership and Communication Theory and Exploratory Study,” Journal of Computer Information Systems, vol. 41, no. 1, February, 2016, pp. 66-78. [68] C. Sauer, C. Cuthbertson, “The State of IT Project Management in the UK 2002-2003,” Computer Weekly, pp. 1-82, 2003. [69] R. Schmidt, K. Lyytinen, M. Keil, P. Cule, “Identifying Project Software Risks: An International Delphi Study,“ Journal of Management Information Systems, vol. 17, no. 4, March, 2001, pp. 5-36. [70] I. Kaya, M. A. Oner, A. N. Basoglu, Critical Success Factors in R&D Project Management in Military Systems Acquisition and a Suggested R&D Project Selection Methodology for Turkish Armed Forces [Online]. Available: http://www.maoner.com/2003_16_08CSFsR%26D.PDF [71] G. Lally, Understanding Information Technology System Project Failure. Dublin Institute of Technology, School of Computing Research Paper (ITSM) [Online]. Available: https://pdfs.semanticscholar.org/c39e/c6deeeaf3baa63854c1df090ffddae960a4c.pdf?_ga=2.59108462.278003 638.1564064101-1773118598.1562166338 [72] J. Jiang, G. Klein, J. Balloun, „Ranking of system implementation success factors,“ Project Management Journal, vol. 27, no. 4, December, 1996, pp. 49-54. [73] S. H. Hong, L. S. Eun, H. R. Choi, K. Lee, M. J. Cho, “A Study on Critical Success Factors of Korean Government-led RFID/USN Projects,” International Journal of Advancements in Computing Technology (IJACT), vol. 4, no. 3, February, 2012, pp. 279-284. [74] M. H. N. Nasir, S. Sahibuddin, „Critical Success Factors for Software Projects: A Comparative Study,“ Scientific Research and Essays, vol. 6, no. 10, April, 2011, pp. 2174-2186. [75] J. S. Reel, “Critical Success Factors in Software Projects,” IEEE Software, vol. 16, no. 3, May-June, 1999, pp. 18-23. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 [76] J. A. Ogden, “Supply base reduction: an empirical study of critical success factors,“ The Journal of Supply Chain Management, vol. 42, no. 4, October, 2006, pp. 29-39. [77] H. Drummond, “Riding a tiger: some lessons of Taurus,” Management Decision, vol. 36, no. 3, April, 1998, pp. 141-146. [78] P. Handerson, Why Large IT Projects Fail [Online]. Available: http://www.ecs.soton.ac.uk/˜ph/LargeIT.pdf [79] N. G. Levenson, “The Role of Software in Spacecraft Accidents,“ AIAA Journal of Spacecraft and Rockets, vol. 41, no. 4, July, 2004, pp. 564-575. [80] B. Nuseibeh, “Ariane 5: Who Dunnit?,“ IEEE Software, vol. 14, no. 3, pp. 15-16, May-June, 1997. [81] P. Beynon-Davies, „Human Error and Information System Failure: The Case of the London Ambulance Service Computer-Aided Despatch System Project,“ Interacting with Computers, vol. 11, no. 6, June 1999, pp. 699-720. [82] J. Jiang, G. Klein, “Software Development Risks to Project Effectiveness,” Journal of Systems and Software, vol. 52, no. 1, May, 2000, pp. 3-10. [83] S. Hastie, What Makes Information Systems Projects Successful?, Software Education Associates Ltd, May, 2006. [84] R. R. Young, “Twelve Requirements Basics for Project Success,” CrossTalk: The Journal of Defense Software Engineering, vol. 19, no. 12, December, 2006, pp. 4-8. [85] S. Snedaker, R. Rogers, IT Security Project Management Handbook, 1st ed. Syngress Publishing, Inc., 2006. [86] R. S. Frey, Successful Proposal Strategies for Small Business: Using Knowledge Management to Win Government, Private-Sector, and International Contracts, illustrated ed. Artech House, 2002. [87] Q. M. Nguyen, “Planning in Software Project Management, An Empirical Research Software Companies in Vietnam,” PhD. Dissertation, Faculty of Economics and Social Sciences, University of Fribourg, Fribourg, Switzerland, 2006. [88] J. Wan, R. Wang, “Empirical Research on Critical Success Factors of Agile Software Process Improvement,” J. Software Engineering & Applications, vol. 3, no. 12, January, 2010, pp. 1131-1140. [89] E. Oz, “When Professional Standards are Lax: The CONFIRM Failure and its Lessons,“ Communications of the ACM , vol. 37, no. 10, October, 1994, pp. 29-36. [90] J. Glaser, (2004) “Management’s Role in IT Project Failures,” Healthcare Financial Management, vol. 58, no. 10, November, 2004, pp. 90-92. [91] N. Gorijan, L. Ma, M. Mittinty, P. Yarlagadda, Y. Sun, „A review on degradation models in reliability analysis,“ Proceedings of the Fourth Word Congress on Engineering Asset Management, Athens, September, 2009, pp. 28-30. [92] O. Zwikael, J. Smyrk, Project Management for the Creation of Organizational Value, 2011 ed. London, England: Springer, 2011. [93] R. C. Mahaney, A. L. Lederer, “Information System Project Management: An Agency Theory Interpretation,” Journal of Systems and Software, vol. 68, no. 1, pp. 1-9, October, 2003. [94] O. I. Tukel, W. O. Rom, “Analysis of the Characteristics of Projects in Diverse Industries,” Journal of Operations Management, vol. 16, no. 1, January, 1998, pp. 43-61. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020218 [95] W. Belassi, O. I. Tukel, „A new framework for determining critical success/failure factors in projects,“ International Journal of Project Management, vol. 14, no. 3, June, 1996, pp. 141-151. [96] M. Hirshfield, P. D. Lee, Critical Success Factors for Healthcare Software Implementations, Vitalize Consulting Solutions, inc., 2006. [97] S. Appelbaum, A. Steed, “The critical success factors in the client-consulting relationship,” Journal of Management Development, vol. 24, no. 1, January, 2005, pp. 68-93. [98] T. K. Perkins, “Knowledge: The Core Problem of Project Failure,“ Crosstalk: The Journal of Defense Software Engineering, vol. 19, no. 6, June, 2006, pp. 13-15. [99] V. Kuo, A Study on the Types of Managerial Behaviors, Styles and Practices that Lead to Project Success, Massachusetts Institute of Technology, 2006. [100] R. J. Turner, The handbook of project-based management, 3rd ed. McGraw-Hill Professional, 2008. [101] D. Tesch, T. J. Kloppenbourg, M. N. Frolick, “IT Project Risk Factors: The Project Management Professionals Perspective,” Journal of Computer Information Systems, vol. 47, no. 4, June, 2007, pp. 61-69. [102] S. C. Misra, V. Kumar, U. Kumar, “Different Techniques for Risk Management in Software Engineering: A Review,” in ASAC Conference in Banff, Alberta, Canada, 2006, pp. 196 – 205. [103] T. Julli, Leadership Princpiles for Project Success, 1st ed. Boca Raton, Florida: CRC Press, 2010 � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Information Systems Project Success Factors: Literature Review Author Author Nermina Durmic Abstract A summary of the resource. The purpose of this paper is to identify and collect most commonly discussed project success factors in the context of information systems (IS) projects. Through the process of review of 88 books, relevant studies and scientific works 72 success factors were detected, with a total of 689 appearances, which are then classified into six factor groups: Planning, Project team, Project management, Development, Customer, Project facilitation. The paper reveals that factors that were recognized as the most critical ones for the success of information systems projects by majority of authors belong first to Planning, and then to Project team and Project management groups of factors. Findings in this paper are expected to serve as a valuable theoretical basis for future empirical research of success and failure of projects in modern information technologies (IT) organizations, and development of related IS project success models. Keywords Keywords. Project success, project success factors, information systems, literature review. Identifier An unambiguous reference to the resource within a given context 2637-2835 Publisher An entity responsible for making the resource available International Burch University, Sarajevo, Bosnia and Herzegovina Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering Date A point or period of time associated with an event in the lifecycle of the resource January, 2020 https://eprints.ibu.edu.ba/files/original/097f70db222c171be9a1b8c5a393af84.pdf b3ead4db3fc460f477e7d5dce620b6e5 PDF Text Text Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 Analysis of Transient and Voltage Stability of an 11-Busbar Testing System Alma Halilović1, Mirza Šarić1 1 International Burch University, Sarajevo, Bosnia and Herzegovina alma.halilovic@stu.ibu.edu.ba mirza.saric@ibu.edu.ba Abstract – A comprehensive treatment from the physical and mathematical perspective supply modelling, analysis, control and covers a range of topics including modelling, computation of load flow in the transmission grid, stability analysis of the transient state. It is widely accepted that transient stability is an important aspect in designing and upgrading electric power system. In utility planning, transient stability is studied by numerical simulation. It involves the study of the power system following a major disturbance. In order to study Electric Power System transient stability, the models to describe their components should be defined. The components are defined using the classical model, which is valid to time periods up to 2 seconds. This project contains 11 busbars, 1 synchronous generator, 3 loads and 8 transformers. This research is done in DIgSILENT PowerFactory software for network modeling and simulation by using Stability Analysis Functions (RMS) advanced feature. In this paper we are analyses the maximum rated power of distributed generation (DG) considering only the terms of voltage limit constraints, the N-1 operational criterion analysis and three-phase symmetrical fault analysis for N-1 criterion is examined. Keywords - high load, n-1 criterion, synchronous generator, transient stability analysis 1. Introduction Stability in Power Systems is one of the importances of a system that had increased. It is the most widely used from power blackouts. Today, usage of power systems interconnection had increased using of new technologies and controls, and the increased its usage in highly demanding situations. In order to maximize the system stability research on power system stability should be carried out. In order to design the perfect system to solve this problem a detailed study of the design should be performed. In the last years [1], due to the spread of electric generation facilities and economic factors, Electric Power Systems operate more closely to their limits. Thus, more than before, it is of crucial importance the existence of methods to assess the system stability. In [2] there are two kinds of stability problems: voltage stability and transient stability. This paper addresses the transient stability. Transient stability analysis of a power system is concerned with the system’s ability to remain in synchronism following a disturbance. Following a large disturbance, the synchronous alternator the machine power (load) angle changes due to sudden acceleration of the rotor shaft. The objective of the transient stability study is to ascertain whether the load angle returns to a steady value following the clearance of the �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 disturbance. The loss of synchronism develops in a very few seconds after the disturbance inception, among the phenomena transient stability is the fastest to develop. These challenging aspects motivate our choice to mainly concentrate on transient stability. The conventional transient stability measure of the system robustness to withstand a large disturbance is its corresponding critical clearing time (CCT). This is the maximum time duration that the disturbance may act without the system losing its capability to recover a steady-state operation. Another transient stability measure of great practical importance is the power (generation or transfer) limit. It is defined as the largest power sustainable without loss of synchronism, given the occurrence of a large disturbance and it’s clearing scenario. In [4], note that the measures imply consideration of three distinct phases: the pre-fault, the during-fault, and the post-fault one. In [5] faults need to be cleared within critical clearing time and after that system need to be able to regain stability. Some of the most important parameters influencing stability are fault clearance time, fault location, and type of the fault. 2. Literature Review Mania Pavella [3] identified three classes of approaches – Decision Tree, KNN and neural network - to transient stability and analyzed which can meet most stringent requirements of transient stability. It is found that stability could be achieved with the appropriate combinations of numerical, direct, and automatic learning techniques. Mirza Saric and Irfan Penava [5] in their research discussed theoretical background of induction generator, its simulation model, as well as dynamic response analysis procedure for a wind farm connected to real network. Thought their research they showed the importance of transient stability in case of integration of large renewable sources to the network. In terms of rotor angle, frequency and voltage stability issues the observed case of wind farm integration was not appropriate to connect to the network with induction generator as the rotor speed was too large, with sharp reduction of reactive power as voltage and active power equal to zero for period that are too long for system to operate in stable state. Innocent Davidson and Immanuel Mbangula [6] examined and analyzed the fault that appeared on the 330kV transmission line between Omburu sub-station and Ruacana power station where the blackout happened for 6 hours. The goal of this research was to investigate what fault occurred, what is the cause and solutions to prevent such fault occur again. The results from DIgSILENT PowerFactory are compared with data obtained from NamPower records and it is found that it was the single phase to ground fault. Ioanna Xyngi, Anton Ishchenko, Marjan Popov, and Lou van der Sluis [7] in their research described the transient stability analysis of a 10-kV distribution network with wind generators, microturbines, and CHP plants modeled in Matlab/Simulink and investigated faults that are simulated on various locations. They showed that in the network with distributed generators (DGs) the protection settings must be adjusted accordingly in order to have stable system as the undervoltage protection should be different for different DGs. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 Diaz-Alzate, Candelo-Becerra and Villa Sierra [8] investigated and found a new way of managing and controlling transient stability based on relative angles. They showed how predefined thresholds of relative angles which they attained by offline simulations and the relative angles attained during the online operation with PMUs are useful in the process of monitoring as well as predicting transient stability under real-time operation. They performed analysis on New England with 39 busbars and IEEE with 118 busbars networks with different contingencies and control actions that are applied at predicted time. The relative angle and the predefined thresholds of the relative angle helped in monitoring and predicting of the systems instability with enough time to respond to oscillations that appeared in the system. 3. Methodology In this part we will show 3 different steps of methodology. In first step we will search and show what is the maximum power that satisfies standards of synchronous generator while checking the voltage profile. In the second step we will do N-1 operational criterion analysis for 6 different cases in this grid and found the new maximum power that satisfies the network. In the last step we will do three phase symmetrical fault analysis for N-1 criterion, with 4 different cases in it. (a) In the first part of the project the maximum rated power of a generator at BB13 (DG), considering only the voltage limit constraints is found, and then the three-phase symmetrical fault analysis is performed for the value of maximum power. We investigate fault duration period of 0.02 seconds. After the RMS simulation is done, voltage development, rotor angle, active and reactive powers are plotted for the further analysis. b) In the second step the n-1 criterion in terms of voltage profile development, as well as line and transformer loading, for 6 different cases for high load scenario, with maximum DG power was investigated. An operational criterion N-1 can be applied to the existing network, if there may be planned or unplanned congestion that may exist at a certain moment. If the criterion is satisfied, that means the system will be able to support a predefined contingency, operating after that, with a minimum performance. The N-1 criterion requires that the system can be able to tolerate the outage of any one component without disruption and does not concern itself with the probability of an outage. If an outage is highly unlikely, the criterion is still generally applied because system failure due to a lost component is unacceptable. The criterion is generally considered as the need to balance generation and load. For modeling network for 6 different scenarios, 6 lines from which the network is consisted are modeled with the following event that placed lines one by one out of service. The voltage on the critical busbar – point of coupling - was monitored along the way, as well as the voltage in the entire system, with the line and transformer loading, making sure that the voltage or line and transformer loading does not exceed the permitted limit specified by the standard EN 50160 for delivered power quality. Minimum upper limit for the voltage, as denoted in the EN 50160 standard, is 0.9 p.u., while maximum upper limit for the voltage is 1.1 p.u. Regarding the line and transformer loading the constraint limits determined by the standard EN 50160 is that it does not exceeds loading of 100 %. In such way the n-1 criterion is investigated and the maximum power for which the system is in the stable mode operation is found for each of the different scenarios when one by one line were out of service. For different events that �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 placed different line out of service, the system will be stable for different maximum power produced from generator. c) In the next part of the research the three-phase symmetrical fault is simulated and the rotor angle, active power, reactive power and voltage development for the following cases, for high load conditions and maximum DG power is to be reported in the results and discussion part of this research project. Three-phase symmetrical fault is simulated in the DIgSILENT PowerFactory software for network modeling and simulation, one second before, during and 5 seconds following the fault on 6 different lines for each of the case scenario. This research is done in DIgSILENT PowerFactory software for network modeling and simulation by using Stability Analysis Functions (RMS) advanced feature. First, the maximum power generation from synchronous generator is found by using the power flow option in the program where all voltage on all busbars, as well as line and transformer loading in the network, is checked so to satisfy EN 50160 standard limits. Next, in order to satisfy limits, set by EN 50160 standards in terms of voltage profile, line and transformer loading when 6 lines of which the system is consisted are modeled with an event that placed line one by one out of service. In such way, by using the power flow analysis function when lines are out of service the n-1 criterion is investigated. For the third part of the research, where three-phase symmetrical faults are simulated on each of the 6 lines that were investigated. The duration of the fault is for the analysis of the 4 case scenarios set on lines to be 0.02, 0.05, 0.5 and 5 seconds and then rotor angle, active power, reactive power and voltage development for the following cases, for high load conditions and maximum DG power are plotted in the graphs, and in such way the behavior of the system could be tracked. d) In the last part of this project the power factor is changed when line 2 is followed with an event that placed it out of service, thus when investigating the N-1 criterion. Values of power factor that are taken into consideration are 0.8, 0.9 and 1, both capacitive and inductive while the voltage development is analyzed and new maximum power of generator is determined in order to satisfy EN 50160 standard for voltage variation in the network for new values of power factor. 4. Results and Discussion A. Maximum Power from Generator in Terms of Voltage Profile In the first part the maximum voltage in the network is found by changing the power generated by the synchronous generator while checking the voltage profile of the network in order that voltage does not increase beyond allowed limits. In the Table III. main results of power flow in terms of voltage magnitude and angle at the busbar BB11 which is the point of coupling, are shown when network operates with no any faults occurring in the system. Maximum power generated from synchronous generator that satisfy the voltage limits set by EN 50160 standard is found to be 286 MW, while the maximum power that satisfy all criterions from the standard – voltage, line and transformer loading – is found to be 104 MW. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 Fig. 1. Network Model from DIgSILENT PowerFactory software The power stability analysis of the given network is done in terms of three-phase symmetrical fault analysis for the maximum power which satisfies the allowed voltage limits and equals to 286 MW. The three cases are investigated where the duration of the fault is set to be 0.02 seconds in fist case, 0.05 seconds in second case and 0.2 seconds in third case. The results obtained when duration of fault is taken to be 0.02, which is expected to show the best results due to lowest fault duration, are shown in Fig.1. In the Fig 1. where the graphs of voltage of the busbar which is the point of coupling, angle of rotor of the synchronous generator, as well as active and reactive power of the synchronous generator are plotted. It can be seen from the graphs that for the maximum power that satisfy the EN 50160 standard only in terms of voltage profile, the system is unstable even for the very small periods of fault duration and it shows that for this high power of generator the network quite inadequate to operate. B. N-1 Operational Criterion Analysis Summary of results obtained when different lines in the network placed out of service are presented in the Table I. and dynamic response of voltage profile in Fig.2., as well as rotor angle in Fig.3. It can be observed which line is the most critical and which is maximum power for such line. The maximum power that satisfies the n-1 criterion for the entire network is found to be 93 MW, since it satisfys the n-1 criterion for the most critical line, which is little bit lower than the maximum power that satisfys the EN 50160 standard for delivered power quality when all lines are in service that is found to be 104 MW. The maximum power for line 1 and line 2 is 93 MW, thus, this value of generated power is compared with maximum power of all other lines in the system, while the voltage magnitude, as well as angle, does not change significantly. In the Table II. the line and transformer loading is compared for same lines for their maximum power and value of 93 MW, and as it can be seen, there is approximately more that 10% of difference in line loading and around 4% in transformer loading. From the results obtained it can be concluded that the most realiable solution is to consider value 93 MW as maximum power that satisfy N-1 criterion for entire network. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 Table 1. Voltage developments for maximum power for each line in the network determined by N-1 criterion Out of service (BB11) Line 1 93 MW Line 2 93 MW Line 3 93 MW Line 3 105 MW Line 4 93 MW Line 4 105 MW Line 5 93 MW Line 5 100 MW Line 6 93 MW Line 6 105 MW UI, Magnitude MV 67,81801 67,1283 67,4659 68,03866 67,39707 67,94841 65,03935 65,13477 67,48871 68,16291 u, Magnitude p.u. 1,027546 1,017095 1,022211 1,030889 1,021168 1,029521 0,985445 0,98689 1,022556 1,032771 U, Angle deg 6,905885 6,473696 6,748108 8,197442 6,682203 8,111645 8,237019 9,947232 5,909104 7,01058 Table 2. Line and transformer loading for N-1 criterion in % N-1 criteria (BB11) Line 1 Line 2 Line 3 Line 4 Line 5 Line 6 Power Loading [%] line 2/1 93 MW 93 MW 93 MW 105 MW 93 MW 105 MW 93 MW 100 MW 93 MW 105 MW 97,9854 99,39992 87,84784 99,58101 87,93902 99,71524 91,17996 98,65325 87,81764 99,39667 Loading [%] Two-winding transformer (5) 93,85382 96,90187 81,66314 84,96854 81,74933 85,08491 84,81535 87,13317 81,63461 84,80878 Fig. 2. Dynamic response for 0.02 s fault duration on lines – Voltage Fig. 3. Dynamic response for 0.02 s fault duration on lines – Rotor Angle �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 C. Three Phase Symmetrical Fault Analysis for N-1 Criterion Since it has been shown that the system is unstable for the maximum power that satisfy EN 50160 standard in term of voltage variation, the three-phase symmetrical fault analysis is done after the N-1 criterion is satisfied for the observed network. The maximum power on the most critical line is previously found to be 93 MW by using the N-1 criterion, and in this part, this is the power for which the system is analyzed. The four cases are examined where the duration of the fault is set to be 0.02 seconds in the first, 0.05 seconds in the second, 0.5 seconds in the third and 5 seconds in the fourth case. Since the line 2 is shown to be most critical in the network after the N-1 criterion analysis done for this case, the result for this line are shown in the graph plotted after the RMS simulation done. In the Fig. 4. the dynamic response for 0.02 seconds fault duration on line 2 in terms of voltage profile is shown. As it can be seen from the graph, when the fault occurs in the system on the line 2, there is voltage drop at the time of fault occurrence and as the duration of the fault is 0.02 seconds, the system returns to be in the balance eventually. In the Fig. 5. there is rotor angle of synchronous generator, active and reactive power presented. As it can be observed from the figures, the system is for the 0.02 seconds fault duration stable and the network for power of 93 MW generated from synchronous generator is adequate and secure for operation as the system after some period reaches equilibrium state. The dynamic response in terms of voltage profile is same as for 0.02 s which is shown is Fig.3., while in Fig. 6. dynamic response of rotor angle of synchronous generator, active as well as reactive power is shown when duration of fault is taken to be 0.05 s for the three-phase symmetrical fault simulation. As it can be seen from Fig. 5. when the fault occurs in the system on the any line, there is voltage drop as the fault occurs on the line, after which, for all lines except line 1, system is stable and able to reach new equilibrium. Voltage dynamic response is same as in Fig.3 when the fault occurs in the system on line 1. There is the voltage drop where voltage drops to approximately zero for longer period than it was case on all other lines, and it will oscillate in the range from value slightly higher than 0 p.u. to approximately 0.35 p.u., with no signs that it will eventually come to the state of balance. However, observing the dynamic response of rotor angle, active and reactive power for the line 1 from Fig.7. for 0.05 s fault duration, the system is unstable to adequately operate since it is unable to attain equilibrium state again. In the Fig. 8. and Fig. 9. after the RMS simulation is done, the results obtained show that in terms of voltage development seen from the Fig. 8., at the time of fault occurrence there appears the voltage drop where the system goes back to the balanced state eventually. In the Fig. 9. the system response of rotor angle of synchronous generator, active and reactive power show that the system for the 0.5 seconds fault duration is stable and the network for power of 93 MW generated from synchronous generator is adequate and secure for operation in case for all analyzed lines except for line 2. Considering case when fault simulated on line 2 for 0.5 s, the dynamic response of voltage development, rotor angle, active and reactive power is similar as response in Fig. 8. while voltage dynamic response is shown in Fig. 10. There is a major voltage drop in the system which is the consequence of the fault that occurred on the line 2. The voltage, after it drops to value slightly higher than zero, is oscillating from that value to approximately 0.35 p.u. Since, also, all other �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 parameters continued to oscillate equally for all 10 seconds and t will not came back to the balanced state, thus, the system is denoted as unstable in terms of all parameters examined. Fig. 4. Dynamic response for 0.02 s fault duration on line 2 – Voltage Fig. 5. Dynamic response for 0.02 s fault duration on line 2 – Rotor Angle (blue), Active (purple) and Reactive Power (green) Fig. 6. Dynamic response for 0.05 s fault duration on lines – Rotor Angle (blue), Active (purple) and Reactive Power (green) In the Fig. 11. the results obtained from the three-phase symmetrical fault analysis when duration of fault is taken to be 5 seconds occurring on the line 2 is shown in terms of voltage profile and in Fig. 12. rotor angle response is shown. Since the duration of the fault on lines is taken to be 5 seconds, which represents too long period for a fault, it can be observed that there is the breakdown of the system in terms of rotor angle where the dynamic response is infinitely oscillating where the speed of the rotation is increasing causing system to �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 be out of balance. There is with the voltage drop at the time the fault occurs there is rise in the angle of rotor which starts to excessively rotate, and drop in active power to the respect of reactive power rise. At the time there is fault on the line 2 it can be observed that all analyzed parameters are oscillating in the smaller range, while at approximately 6th second, when the duration of the fault is over, there are greater oscillations where the system is unable to reach the equilibrium state and, instead, and reaches the state of complete breakdown. Fig. 7. Dynamic response for 0.05 s fault duration on line 1 – Rotor Angle (blue), Active (purple) and Reactive Power (green) Fig. 8. Dynamic response for 0.5 s fault duration on lines – Voltage Fig. 9. Dynamic response for 0.5 s fault duration on lines – Rotor Angle (blue), Active (purple) and Reactive Power (green) �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 Fig. 10. Dynamic response for 0.5 s fault duration on line 2 –Voltage Fig. 11. Dynamic response for 5 s fault duration on lines – Voltage Fig. 12. Dynamic response for 5 s fault duration on lines – Rotor Angle D. Voltage Control by Power Factor Correction for N-1 Criterion In the Table III. The results of voltage development on the BB11 busbar are presented when the power factor of the synchronous generator is varied. As the generator is operated under different power factor, there is a change in the active and reactive power production in order to maintain voltage within limits set by EN �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 50160 standards. The results obtained show how the voltage is regulated by power factor correction as the in the case of 0.8 capacitive power factor voltage magnitude for 93 MW falls below allowed limits, and new maximum power is determined to be 45 MW. As in the case of 0.8 inductive power factor the voltage goes beyond upper allowed limit in case when power production of 93 MW, and the new maximum power for stable operation is found to be 74 MW. In cases when power factor is 0.9 in both capacitive and inductive mode, the voltage on the critical busbar, as well in the entire network, stays within allowed limits and does not change significantly for both power of 93 MW and new found power. It can be then concluded that if the generator operates under power factor of 0.8 in both modes, there is less active power produced, while if it operates under power factor of 0.9 in both modes, as well as under power factor of 1, the maximum power of 93 MW previously determined satisfy N-1 criterion and offers stable and secure network operation. Considering results obtained and presented in Table IV., in terms of line and transformer loading, for power of 93 MW any power factor in either capacitive or inductive mode less than unity, the power of 93 MW will distort the network operation and stability. Thus, it can be concluded that considering only voltage constraints the system will be stable for power factor of 0.9 capacitive and inductive, while when line and transformer loading taken into consideration, there is all cases when power factor less than 1 new maximum power for unstable and unsecure network operation. Table 3. Voltage developments for different power factor Power factor BB11 Line 2 0.8 (45 MW) Line 2 0.8 (83 MW) Line 2 0.8 (74 MW) Line 2 0.8 (93MW) Line 2 0.9 (73MW) Line 2 0.9 (93MW) Line 2 0.9 (89MW) Line 2 0.9 (93MW) Line 2 1 (93MW) CAP CAP IND IND CAP CAP IND IND IND UI, Magnitude MV 59,59856 55,44239 72,79612 75,16263 60,93922 60,2638 72,12859 72,52091 67,06188 u, Magnitude p.u. U, Angle deg 0,903006 0,840036 1,102971 1,138828 0,923322 0,913088 1,092857 1,098802 1,016089 3,5407772 13,12576 1,038122 2,164138 7,032856 10,3019 3,275636 3,567566 6,509941 Table 4. Line and transformer loading for different power factor Power Factor CAP CAP IND IND CAP CAP IND IND IND 4. Line 2 0.8 Line 2 0.8 Line 2 0.8 Line 2 0.8 Line 2 0.9 Line 2 0.9 Line 2 0.9 Line 2 0.9 Line 2 1 Power 45MW 93MW 74MW 93MW 73MW 93MW 89MW 93MW 93MW Loading [%] line 2/1 40,80236 167,4764 92,47962 116,5894 98,69827 131,5861 99,38897 104,098 99,53208 Loading [%] Twowinding transformer 74,93822 123,1931 87,84803 98,82892 93,07358 109,3657 93,99311 96,17869 96,96105 Conclusion Voltage stability is the main problem concerning utilities due to the continuous growth and deregulation. In this paper, the network with 11 busbars is examined for a transient stability. Transient stability studies deal �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020217 with the effects of large and sudden disturbances that occurs within the network such as it is a fault, the sudden outage of a line or the sudden removal or application of load. It is very important to do the transient stability analysis of a system in order to ensure that the system can handle the transient condition which is followed by a major disturbance. Transient stability and voltage instability analysis done in this research for the network with high load conditions shows that for a smaller period of fault duration, the system can be denoted as stable, while in case when the fault duration is somewhat higher than 0.02 seconds, the system is unstable and not secure in operation being unable to attain the equilibrium state and synchronism considering all parameters analyzed – voltage development, rotor angle speed, active and reactive power. In cases when fault duration greater than 0.02 seconds, there is breakdown state appearing in the network after the fault where the rotor speed is increasing, thus, causing system to go unstable and out of synchronism. Also, during the fault, there is voltage drop, active power drop and reactive power rise, after which there is an oscillation which cannot reach equilibrium. When fault duration increased to 0.05 or 0.5 seconds, the system is unstable only when fault simulated on one line, while when fault duration is increased to 5 seconds, then for every line examined and analyzed for three-phase symmetrical fault, the system is unstable. REFERENCES [1] Kundur, P. (1994) Power System Stability and Control. USA: McGraw-Hill, Inc. [2] Sucena P., J. P. (2007) Redes de Energia Eléctrica: uma análise sistémica. 2ª edição. Lisboa, IST Press. [3] Mania P. (1997) Power System Stability and Control Comparative Analysis. IFAC Control of Industrial Systems, Belfort, France. [4] Lau, Mark A. and Kuruganty, Sastry P. (2010) "A Spreadsheet Illustration of the Transient Stability Analysis of Power Systems," Spreadsheets in Education (eJSiE): Vol. 3: Iss. 3, Article 6. Available at: http://epublications.bond.edu.au/ejsie/vol3/iss3/6 [5] Saric, M., & Penava, I. (2014, May). Transient stability of induction generators in wind farm applications. In 2014 14th International Conference on Environment and Electrical Engineering (pp. 230-235). IEEE. [6] Davidson I. and Mbangula I. (2014) Power System Modelling and Fault Analysis of NamPower’s 330 kV HVAC Transmission Line. Journal of Energy and Power Engineering 8 (pp. 1432-1442). [7] Xyngi I., Ishchenko A, Popov M., and van der Sluis L. (2009) Transient Stability Analysis of a Distribution Network With Distributed Generators. IEEE Transactions on Power Systems, vol. 24, no. 2. [8] Diaz-Alzate, A. F., Candelo-Becerra, J. E., & Villa Sierra, J. F. (2019). Transient Stability Prediction for Real-Time Operation by Thresholds. Energies, 12(5), 838. Monitoring the Relative Angle with Predefined � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Analysis of Transient and Voltage Stability of an 11-Busbar Testing System Author Author Alma Halilović, Mirza Šarić Abstract A summary of the resource. A comprehensive treatment from the physical and mathematical perspective supply modelling, analysis, control and covers a range of topics including modelling, computation of load flow in the transmission grid, stability analysis of the transient state. It is widely accepted that transient stability is an important aspect in designing and upgrading electric power system. In utility planning, transient stability is studied by numerical simulation. It involves the study of the power system following a major disturbance. In order to study Electric Power System transient stability, the models to describe their components should be defined. The components are defined using the classical model, which is valid to time periods up to 2 seconds. This project contains 11 busbars, 1 synchronous generator, 3 loads and 8 transformers. This research is done in DIgSILENT PowerFactory software for network modeling and simulation by using Stability Analysis Functions (RMS) advanced feature. In this paper we are analyses the maximum rated power of distributed generation (DG) considering only the terms of voltage limit constraints, the N-1 operational criterion analysis and three-phase symmetrical fault analysis for N-1 criterion is examined. Keywords Keywords. high load, n-1 criterion, synchronous generator, transient stability analysis Identifier An unambiguous reference to the resource within a given context 2637-2835 Publisher An entity responsible for making the resource available International Burch University, Sarajevo, Bosnia and Herzegovina Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering Date A point or period of time associated with an event in the lifecycle of the resource January, 2020 https://eprints.ibu.edu.ba/files/original/6f132e73da7a55449a2406e39a5fcb3b.pdf d10a56d18bf5288213375abc8cf94ffe PDF Text Text Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020216 Distribution of inherited thrombophilia markers in Bosnian-Herzegovinian population: a review of previous studies Nermin Đuzić1, Adna Ašić1 1 International Burch University, Sarajevo, Bosnia and Herzegovina nermin.djuzic@stu.ibu.edu.ba adna.asic@ibu.edu.ba Abstract – Thrombophilia is a condition that is associated with an individual’s risk for venous or arterial thrombosis, as well as a risk of adverse pregnancy outcomes. Gene variants that are the most commonly associated with inherited thrombophilia are F5 mutation 1691G>A (Factor V Leiden), F2 20210G>A (prothrombin mutation), MTHFR 677C>T, and PAI-1 variant 4G/5G. This paper aims to review currently available literature on the prevalence of heritable thrombophilia genetic markers and their association with thromboembolic events in Bosnia and Herzegovina. PubMed and PubMed Central databases of the National Center for Biotechnology Information (NCBI) and ResearchGate were searched to identify the most relevant studies. The results of the previously published studies show discrepancies when it comes to reported findings, thus implying that further research on this topic is necessary. It is suggested that new studies include greater sample size in order to confirm the correlation between the studied variants and conditions associated with heritable thrombophilia in the Bosnian-Herzegovinian population and to advance the understanding of these variants. Keywords - Bosnia and Herzegovina, Factor V Leiden, Inherited thrombophilia, MTHFR, PAI-1, Prothrombin mutation 1. Introduction Thrombophilia is a condition putting an individual at risk of venous or arterial thrombosis and is mainly divided into hereditary and acquired. Hereditary thrombophilia is associated with genetic mutations influencing the level or activity of proteins involved in coagulation cascade and includes both loss-offunction and gain-of-function mutations [1,2]. Three genetic markers are the most commonly studied as indicators of inherited thrombophilia. F5 gene variant 1691G>A, usually termed Factor V Leiden, gives rise to a peptide that is uncleavable by the activated protein C (APC) [3]. According to Inbal and Carp (2007), this mutation is responsible for 3-42% of pregnancy losses [4]. If present in homozygous state, this mutation increases a risk for venous thrombosis up to 50-100-fold. The second most studied variant is F2 20210G>A, also known as prothrombin mutation (PTM), that causes elevated levels of prothrombin and 25-fold increased risk of venous thrombosis as well as pregnancy loss [5]. MTHFR gene variant 677C>T codes for thermolabile variant of protein methylene tetrahydrofolate reductase which is a loss-of-function mutation and causes elevated levels of homocysteine in blood, the condition known as hyperhomocysteinemia, which is in turn considered a risk factor for venous thromboembolism [6-8]. The fourth variant indicative of inherited thrombophilia is 4G/5G in type 1 plasminogen activator inhibitor �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020216 (PAI-1) gene, which is a 4328G>T missense variant located 675 bp from the promoter that results in four or five guanine nucleotides in a row [9,10]. The protein product of this gene has a role in fibrinolysis and is associated with adverse pregnancy outcomes, such as intrauterine fetal death, intrauterine growth restriction, preeclampsia, recurrent miscarriage and placental abruption [11]. The aim of this paper is to provide a comprehensive review of the current knowledge regarding the prevalence of heritable thrombophilia markers and their correlation with thromboembolic events in Bosnia and Herzegovina (B&H) based on previously published population studies. 2. Methods In order to investigate associations between the abovementioned genetic variants and inherited thrombophilia that may lead to adverse primary outcomes, National Center for Biotechnology Information (NCBI) databases PubMed and PubMed Central (PMC), and ResearchGate were searched to discover relevant studies published previously. Included were original research papers published in peer-reviewed journals that matched the search of the following keywords: “Factor V Leiden”, “F5 1691G>A”, “prothrombin”, “F2 20210G>A”, “MTHFR 677C>T”, “PAI-1 4G/5G”, “polymorphism”, “thrombophilia”, “Bosnia and Herzegovina”. In order to enhance the search, Boolean “AND” and “OR” operators were used to investigate the association between two searched terms. 3. Prevalence of genetic markers of inherited thrombophilia in Bosnian-Herzegovinian population The search results offered a total of seven studies related to the topic of the prevalence of genetic markers of inherited thrombophilia in B&H, all of them being conducted as of 2013. The study of Karić and colleagues in 2013 was the first study which deals with prevalence of MTHFR 677C>T polymorphism in Bosnian-Herzegovinian population. They studied 102 men and 105 women, who were unrelated and healthy and originating from south-east B&H. At the time of blood sampling, the study participants ranged between 18 and 84 years with a mean age of 45.62 years. The results have shown that 44.44% were heterozygous and 11.11% were homozygous for the study allele. No significant difference was found in allele and genotype frequencies between male and female participants [12]. The first study aiming at analyzing Factor V Leiden prevalence in B&H was performed in 2013 by Valjevac and colleagues. A group of 67 women with mean age of 58.6 years (range 41 to 75 years) with no previous history of cardiovascular diseases and pregnancy loss was recruited and tested. The study failed to find any mutant allele, thus suggesting that, considering functional importance of this allele, there is a need to conduct more research on that topic, as the results of this study were heavily influenced with the small sample size [13]. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020216 The prevalence of polymorphisms Factor V Leiden, prothrombin mutation and MTHFR 677C>T was studied by Adler and colleagues (2014). The study involved a cohort of 100 healthy unrelated individuals from B&H (82 females and 18 males) with the year range of 24-82 years and the mean of 58.8 years. The analyzed loci were in Hardy-Weinberg equilibrium with the following minor allele frequencies (MAF): 6% for Factor V Leiden, 6% for F2 20210A allele, and 37.5% for MTHFR 677T allele. The authors noted the drawback of a small sample size and imbalanced sex ratio. The study, however, demonstrated the coinheritance of thrombophilia markers, since nine participants had Factor V Leiden and MTHFR 677C>T mutant. Compared to 17 European countries, the prevalence of Factor V Leiden and F2 20210G>A variants was significantly higher in B&H [14]. Another study examined the prevalence of the same three variants and their association with deep venous thrombosis (DVT) in B&H [15]. The study group included 111 thromboembolic patients (59 females and 52 males ranging from 21 to 84 years) and 207 healthy controls (105 females and 102 males ranging from 18 to 84 years) with no history of venous thromboembolism (VTE). When it comes to Factor V Leiden prevalence, 18% of the study group patients were heterozygous and 2.7% were homozygous, while 3.86% of the control group participants were heterozygous for this variant, which a statistically significant difference between groups. Statistically significant difference was also found between men with DVT and the control group, as well as between women with DVT and the control women group. F2 20210G>A variant was detected in 2.7% of the study group patients and was absent in control group, which was not statistically significant. Frequencies of MTHFR C677T alleles and genotypes did not differ significantly between the two groups. Allele frequency and functional significance of Factor V Leiden variant detected in this study was in agreement with earlier studies in Caucasians [16-20]. Also, the results of this study were found consistent with the data from other neighboring countries [21-23]. The absence of functional significance of the studied MTHFR polymorphism was also in line with previously published literature [4,16,24,25]. Finally, the authors of this study found that 14.9% of the patients from the DVT group were compound heterozygotes for Factor V Leiden and MTHFR 677C>T variants, therefore proposing further studies that would aim to analyze whether such genotype combination might be a risk factor for DVT development [15]. A study of Mahmutbegović and colleagues (2017) enrolled 308 women, including 154 women who experienced pregnancy loss as the study group (mean age 33 ± 5.4 years) and 154 women with at least one live-born child and without pregnancy loss as the control group (mean age 31.4 ± 6.7 years) to investigate the correlation between three genetic markers and pregnancy loss. Detected allele frequencies were 3.9% in both study and control groups for Factor V Leiden, 1.9% and 1.6% in the study and control groups, respectively, for prothrombin mutation, and 35.7% and 29.9% in the study and control groups, respectively, for MTHFR 677C>T. Although allele frequencies obtained in this study were in accordance with allele frequencies obtained for other European countries, the authors, however, were not able to find the significant correlation between these three variants and pregnancy loss in Bosnian-Herzegovinian women, which may be due to small sample size of women with three or four pregnancy losses recorded [26]. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020216 The prevalence of Factor V Leiden, prothrombin mutation, MTHFR 677C>T and PAI-1 4G/5G in BosnianHerzegovinian women and their correlation with recurrent pregnancy loss was studied by Jusić and colleagues (2018). The study group included 60 women with two or more consecutive pregnancy losses that occurred before 20th gestation week with the same partner and without history of known causes of pregnancy losses due to chromosomal abnormalities, chronic diseases, or infections. The control group was consisting of 80 women with one or more successful pregnancy outcomes and without any pregnancy complication which could lead to the pregnancy loss. The results have shown that Factor V Leiden and MTHFR 677C>T were proved to correlate with recurrent pregnancy loss, while prothrombin mutation and PAI-1 4G/5G were not found to be significantly associated with the pregnancy loss. Reported allele frequencies were as follows: 7.5% for study and 1.88% for control group for Factor V Leiden, 2.5% for study and 0.63% for control group for prothrombin mutation, 39.17% for cases and 25% for control group for MTHFR 677C>T, and 30% for study and 20% control group for PAI-1 4G/5G. The authors are reporting allele frequencies that were in agreement with previous findings for all study polymorphisms. However, the role of genetic factors for inherited thrombophilia in recurrent pregnancy loss is still a matter of debate, especially when it comes to MTHFR and PAI-1 variants studied. Therefore, the authors are suggesting further studies with larger study and control groups, as well as the need to prevent recurrent pregnancy loss by assessing the status of these variants and calculating individual risk and optimum therapy for each patient [27]. In the most recent study by Ašić and colleagues in 2019, the prevalence of common thrombophilia markers was studied in a population of 130 unrelated healthy Bosnian- Herzegovinians of both sexes, from different age groups, with no recorded history of thrombotic events and originating from different parts of the country. Seven markers most commonly associated with the risk of heritable thrombophilia were investigated, namely Factor V Leiden, F2 20210G>A, MTHFR 677C>T, MTHFR 1286A>C, PAI-1 4G/5G, PAI-1 -844G>A and F13 V35L. Whereas some of these markers were examined in previous studies as described above, this is the first study to include MTHFR 1286A>C, PAI-1 -844G>A and F13 V35L polymorphisms in the population of B&H. The results have shown that the two main thrombophilia markers Factor V Leiden and prothrombin mutation appeared with MAF values of 0.023 and 0.008 respectively. For the remaining four loci, reported MAF values were 0.331 for MTHFR 677C>T, 0.323 for MTHFR 1286A>C, 0.446 for PAI-1 4G/5G, 0.588 for PAI-1 -844G>A, and 0.315 for F13 V35L. This study provides the most extensive population data on the prevalence of main heritable thrombophilia risk factors in B&H and reported allele frequencies were consistent with those reported for other European populations [28]. 4. Conclusion Previously conducted studies in B&H represent a small and rather heterogenous group of studies, including either population studies or case-control studies with the aims to report heritable thrombophilia marker prevalence in the population or their potential association with DVT, pregnancy loss or recurrent pregnancy loss. While initial preliminary studies offered surprisingly low or high MAF values for the most well-known genetic variants Factor V Leiden and F2 20210G>A, later studies reported allele and genotype prevalence �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020216 that is in line with reported data for most European populations. While DVT was found to be positively associated with Factor V Leiden variants [15], the data for obstetric complications are more controversial since two studies reported conflicting results. The first study reported no statistically significant association between Factor V Leiden, F2 20210G>A and MTHFR 677C>T and pregnancy loss [26], while the second one reported a significant increase in mutant allele frequency for Factor V Leiden and MTHFR 677C>T in women with recurrent pregnancy loss [27]. Therefore, it is strongly suggested that further studies assess the functional importance of the most important markers for inherited thrombophilias in B&H by clearly defining study and control groups in order to assess the potential association of these variants with conditions such as (recurrent) pregnancy loss and venous thromboembolism with the goal of assessing the importance of genetic testing in Bosnian-Herzegovinian population and establishing the groundwork for the personalized treatment of inherited thrombophilia and conditions connected to it. In that effort, larger study cohorts including individuals from all parts of the country will be necessary. REFERENCES [1] S. Palta, R. Saroa, and A. Palta, “Overview of the coagulation system”, Indian Journal of Anaesthesia, vol. 58, no. 5, pp. 515-523, 2014. [2] S. M. Stevens, S. C. Woller, K. A. Bauer, R. Kasthuri, M. Cushman, M. Streiff, W. Lim, and J. D. Douketis, “Guidance for the evaluation and treatment of hereditary and acquired thrombophilia”, Journal of Thrombosis and Thrombolysis, vol. 41, no. 1, pp. 154-164, 2016. [3] F. Rodeghiero, and A. Tosetto, “Activated protein C resistance and factor V Leiden mutation are independent risk factors for venous thromboembolism”, Annals of Internal Medicine, vol. 130, no. 8, pp. 643-650, 1999. [4] A. Inbal, and H. J. Carp, “Defects in coagulation factors leading to recurrent pregnancy loss,” in Recurrent Pregnancy Loss: Causes, Controversies, and Treatment, 2nd ed. Boca Raton, FL: CRC Press, 2007, pp. 141-160. [5] R. R. Finan, H. Tamim, G. Ameen, H. E. Sharida, M. Rashid, and W. Y. Almawi, “Prevalence of factor V G1691A (factor V‐Leiden) and prothrombin G20210A gene mutations in a recurrent miscarriage population”, American Journal of Hematology, vol. 71, no. 4, pp. 300-305, 2002. [6] L. Dean, “Methylenetetrahydrofolate reductase deficiency,” in Medical Genetics Summaries [Internet]. Bethesda, MD: National Center for Biotechnology Information, 2012. [7] S. E. Hickey, C. J. Curry, and H. V. Toriello, “ACMG Practice Guideline: lack of evidence for MTHFR polymorphism testing”, Genetics in Medicine, vol. 15, no. 2, pp. 153-156, 2013. [8] N. M. van der Put, F. Gabreëls, E. M. Stevens, J. A. Smeitink, F. J. Trijbels, T. K. Eskes, L. P. van den Heuvel, and H. J. Blom, “A second common mutation in the methylenetetrahydrofolate reductase gene: an additional risk factor for neural-tube defects?”, The American Journal of Human Genetics, vol. 62, no. 5, pp. 1044-1051, 1998. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020216 [9] T. Bezgin, C. Kaymaz, Ö. Akbal, F: Yılmaz, H. C. Tokgöz, and N. Özdemir, “Thrombophilic gene mutations in relation to different manifestations of venous thromboembolism: a single tertiary center study”, Clinical and Applied Thrombosis/Hemostasis, vol. 24, no. 1, pp. 100-106, 2016. [10] J. L. Kujovich, “Prothrombin-related thrombophilia,” in: GeneReviews [Internet]. M. P. Adam, H. H. Ardinger, R. A. Pagon, S. E. Wallace, L. J. H. Bean, K. Stephens, and A. Amemiya, A. Eds. Seattle, WA: University of Washington, 2014. [11] C. J. Glueck, M. J. Kupferminc, R. N. Fontaine, P. Wang, B. B. Weksler, and A. Eldor, “Genetic hypofibrinolysis in complicated pregnancies”, Obstetrics & Gynecology, vol. 97, no. 1, pp. 44-48, 2001. [12] A. Karić, R. Terzić, Z. Jerkić, and J. Mustedanagić-Mujanović, “The frequency of C677T methylenetetrahydrofolate reductase (MTHFR) polymorphism in Southern East Bosnian population”, Journal of Biometrics and Biostatistics, vol. 4, pp. 169-171, 2013. [13] A. Valjevac, B. Mehić, E. Kiseljaković, S. Ibrulj, A. Garstka, and G. Adler, “Prevalence of 1691G> A FV mutation in females from Bosnia and Herzegovina-a preliminary report”, Bosnian Journal of Basic Medical Sciences, vol. 13, no. 1, pp. 31-33, 2013. [14] G. Adler, G. Agnieszka, A. Valjevac, E. Czerska, E. Kiseljakovic, and N. N. Salkic, “Prevalence of genetic prothrombotic risk factors: 1691G>A FV, 20210G>A PT and 677C>T MTHFR mutations in the Bosnian population”, Annals of Human Biology, vol. 42, no. 6, pp. 576-580, 2014. [15] A. Jusić-Karić, R. Terzić, Z. Jerkić, A. Avdić, and M. Pođanin, “Frequency and association of 1691 (G>A) FVL, 20210 (G>A) PT and 677 (C>T) MTHFR with deep vein thrombosis in the population of Bosnia and Herzegovina”, Balkan Journal of Medical Genetics, vol. 19, no. 1, pp. 43-50, 2016. [16] Z. Alfirevic, A. M. Simundic, N. Nikolac, N. Sobocan, I. Alfirevic, M. Stefanovic, and Z. Vucicevic, “Frequency of factor II G20210A, factor V Leiden, MTHFR C677T and PAI-1 5G/4G polymorphism in patients with venous thromboembolism: Croatian case-control study”, Biochemia Medica, vol. 20, no. 2, pp. 229-235, 2010. [17] E. Berge, K. B. F. Haug, E. C. Sandset, E., K. K. Haugbro, M. Turkovic, and P. M. Sandset, “The factor V Leiden, prothrombin gene 20210GA, methylenetetrahydrofolate reductase 677CT and platelet glycoprotein IIIa 1565TC mutations in patients with acute ischemic stroke and atrial fibrillation”, Stroke, vol. 38, no. 3, pp. 1069-1071, 2007. [18] T. Koster, J. P. Vandenbroucke, F. R. Rosendaal, H. de Ronde, E. Briët, and R. M. Bertina, “Venous thrombosis due to poor anticoagulant response to activated protein C: Leiden Thrombophilia Study”, The Lancet, vol. 342, no. 8886-8887, pp. 1503-1506, 1993. [19] L. Meglic, M. Stegnar, T. Milanez, M. Bozic, B. Peterlin, P. Peternel, and Z. Novak-Antolic, Factor V Leiden, prothrombin 20210G→A, methylenetetrahydrofolate reductase 677C→T and plasminogen activator inhibitor 4G/5G polymorphism in women with pregnancy-related venous thromboembolism”, European Journal of Obstetrics & Gynecology and Reproductive Biology, vol. 111, no. 2, pp. 157-163, 2003. [20] H. U. Pauer, T. Voigt‐Tschirschwitz, B. Hinney, P. Burfeind, C. Wolf, G. Emons, and J. Neesen, “Analyzes of three common thrombophilic gene mutations in German women with recurrent abortions”, Acta Obstetricia et Gynecologica Scandinavica, vol. 82, no. 10, pp. 942-947, 2003. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020216 [21] T. Arsov, D. Miladinova, and M. Spiroski, “Factor V Leiden is associated with higher risk of deep venous thrombosis of large blood vessels”, Croatian Medical Journal, vol. 47, no. 3, pp. 433-439, 2006. [22] D. Coen, R. Zadro, L. Honovic, L. Banfic, A. S. Rukavina, “Prevalence and association of the factor V Leiden and prothrombin G20210A in healthy subjects and patients with venous thromboembolism”, Croatian Medical Journal, vol. 42, no. 4, pp. 488-492, 2001. [23] V. Djordjevic, L. J. Rakicevic, D. Mikovic, M. Kovac, P. Miljic, D. Radojkovic, and A. Savic, “Prevalence of factor V Leiden, factor V Cambridge, factor II G20210A and methylenetetrahydrofolate reductase C677T mutations in healthy and thrombophilic Serbian populations”, Acta Haematologica, vol. 112, no. 4, pp. 227-229, 2004. [24] C. Beyan, and E. Beyan, “MTHFR 677 CT gene polymorphism is not a risk factor for venous thromboembolism”, Çukurova Medical Journal, vol. 39, no. 1, pp. 189-190, 2014. [25] A. W. Tsai, M. Cushman, M. Y. Tsai, S. R. Heckbert, W. D. Rosamond, N. Aleksic, N.D. Yanez, B. M. Psaty, and A. R. Folsom, “Serum homocysteine, thermolabile variant of methylene tetrahydrofolate reductase (MTHFR), and venous thromboembolism: Longitudinal Investigation of Thromboembolism Etiology (LITE)”, American Journal of Hematology, vol. 72, no. 3, pp. 192-200, 2003. [26] E. Mahmutbegović, D. Marjanović, E. Medjedović, N. Mahmutbegović, S. Dogan, A. Valjevac, E. Czerska, A. Pawinska-Matecka, A. Madlani, and G. Adler, “Prevalence of F5 1691G> A, F2 20210G> A, and MTHFR 677C> T polymorphisms in Bosnian women with pregnancy loss”, Bosnian Journal of Basic Medical Sciences, vol. 17, no. 4, pp. 309-314, 2017. [27] A. Jusić, D. Balić, A. Avdić, M. Pođanin, and A. Balić, “The association of factor V G1961A (factor V Leiden), prothrombin G20210A, MTHFR C677T and PAI-1 4G/5G polymorphisms with recurrent pregnancy loss in Bosnian women”, Medicinski Glasnik (Zenica), vol. 15, no. 2, pp. 158-163, 2018. [28] A. Ašić, R. Salazar, N. Storm, S. Doğan, W. Höppner, D. Marjanović, and D. Primorac, ”Population study of thrombophilic markers and pharmacogenetic markers of warfarin prevalence in Bosnia and Herzegovina”, Croatian Medical Journal, vol. 60, no. 3, pp. 212-220, 2019. � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Distribution of inherited thrombophilia markers in Bosnian-Herzegovinian population: a review of previous studies Author Author Nermin Đuzić, Adna Ašić Abstract A summary of the resource. Thrombophilia is a condition that is associated with an individual’s risk for venous or arterial thrombosis, as well as a risk of adverse pregnancy outcomes. Gene variants that are the most commonly associated with inherited thrombophilia are F5 mutation 1691G>A (Factor V Leiden), F2 20210G>A (prothrombin mutation), MTHFR 677C>T, and PAI-1 variant 4G/5G. This paper aims to review currently available literature on the prevalence of heritable thrombophilia genetic markers and their association with thromboembolic events in Bosnia and Herzegovina. PubMed and PubMed Central databases of the National Center for Biotechnology Information (NCBI) and ResearchGate were searched to identify the most relevant studies. The results of the previously published studies show discrepancies when it comes to reported findings, thus implying that further research on this topic is necessary. It is suggested that new studies include greater sample size in order to confirm the correlation between the studied variants and conditions associated with heritable thrombophilia in the Bosnian-Herzegovinian population and to advance the understanding of these variants. Keywords Keywords. Bosnia and Herzegovina, Factor V Leiden, Inherited thrombophilia, MTHFR, PAI-1, Prothrombin mutation Identifier An unambiguous reference to the resource within a given context 2637-2835 Publisher An entity responsible for making the resource available International Burch University, Sarajevo, Bosnia and Herzegovina Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering Date A point or period of time associated with an event in the lifecycle of the resource January, 2020 https://eprints.ibu.edu.ba/files/original/6b2e464890bd658d55b645c2200671f3.pdf 368e8f19562326e2eedc5b353024065f PDF Text Text Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020215 Machine Learning in Autism Spectrum Disorder Diagnosis Naida Nalo1, Jasmin Kevrić1 1 International Burch University, Sarajevo, Bosnia and Herzegovina naida.nalo@stu.ibu.edu.ba jasmin.kevric@ibu.edu.ba Abstract— This paper represents an overview of Machine Learning techniques used in Autism Spectrum Disorder - ASD diagnosis. ASD is detected based on behavioral screening which is time consuming and can only be taken by a medical professional. The idea is to find a smaller number of features that are still able to equally well provide satisfying results and not lose the accuracy, sensitivity nor specificity. Some of the algorithms mostly used in recent studies were Artificial Neural Network ANN and Alternating Decision Trees - ADTrees. The researches usually use WEKA software package for applying the algorithm and obtaining results. Keywords—Machine Learning, Autism Spectrum Disorder, diagnosis, features, ANN, ADTree, WEKA. 1. Introduction Autism Spectrum Disorder (ASD) is defined as a developmental disorder that reflects in difficulty to communicate and interact with people, to have minimal interests and to generate pattern like behaviors. There are three separate conditions that are combined into the Autism Spectrum Disorder and those are: Autistic disorder, Asperger’s syndrome and Pervasive developmental disorder not otherwise specified (PDD-NOS). Even though ASD can be a lifelong disorder, an early diagnosis and proper treatment can help improve communications skills and overall ability to function. However, the diagnosis sometimes takes a lot of time, which prolongs the appropriate treatment [1]. Two widely used clinical diagnosis tools for diagnosing autism are The Autism Diagnostic Interview-Revised (ADI-R) [2] and Autism Diagnostic Observation Schedule (ADOS) [3]. ADI-R consists of ninety-three questions that are to be answered by a clinician. This process can take up to two and a half hours to conduct. The ADOS contains four modules, that can be used to test children and adults, according to behavioral and language levels of the person to be tested. This tool uses an algorithm that results with a diagnosis based on the scores of responses. Each module has its own scores [4]. It has become of great importance to find a faster but reliable method of diagnosing ASD, since the earliest treatment gives a greater chance for improvement. �This paper gives insight on the studies conducted in the past on the subject of Machine Learning in ASD diagnosis. A brief review on differences between two versions of Diagnostics and Statistical Manual of Mental Disorders (DSM-IV [5] and DSM-5 [6]) is made. 2. Literature review The ADI-R is one of two most widely used instruments for behavioral diagnosis of ASD [7]. It is structured in a form of an exam containing ninety-three questions and can be applied to individuals from the age of eighteen months and above. The questions are answered by a trained professional but still take up to two and a half hours to finish. And additionally, the gap between the initial screening and the resulting diagnosis can be around thirteen months, depending on the socioeconomic status of the family [8]. This introduces an additional delay in the early treatment crucial for proper development of the person, especially children. In [9], it was proposed to create an exam that can be conducted in minutes, rather than hours and receive satisfying results. Machine learning was used to select the right amount of questions, out of initial ninety-three, that would be able to classify the person in either autism or non-spectrum class. In total, fifteen algorithms were tested, and the one that performed the best with the given data was found to be the Alternating Decision Tree (ADTree). This classification algorithm managed to successfully classify all individuals diagnosed with ASD using only eight questions, that were previously tested with a complete set of ninety-three questions of ADI-R and misclassified only one. However, this research was proved to be unreliable by [10]. This paper brought to attention the importance of understanding both the computational and clinical area before giving any conclusions. The research conducted in [9], was limited by the imbalance of data as well as excluding a big part of it due to missing values. This resulted in only a two-class diagnosis, when originally it should have been three. The middle class, which is the most difficult to identify, that was removed was the ASD Spectrum, leaving only the ASD and Nonspectrum cases. A recommendation from [10] is to use the Unweighted Average Recall – UAR, which is a measure of performance that works better for such unbalanced data and that was used in this paper, when they tried to replicate the work done in [9]. Their results were algorithm dependent and if another algorithm was to be applied to the same data, the number of features would vary. Another important issue that was discussed in [4] were the differences between results of DSM-IV and the new DSM-5. From [4] we learn that although both of the screening methods abovementioned have shown good sensitivity, specificity and high reliability in experiments, the majority of those studies were based on the DSM-IV rather than the new criteria for diagnosing ASD, the DSM-5. Several studies that were mentioned in [4] had conflicting results when using the two different versions of the manual. This introduces the need to reevaluate the current tools for diagnosis, and to adjust them to the new criteria of diagnosing ASD. �Combining all the research and experiments from the past, and critically analyzing the results, suggestions and advices for the future projects are provided in [11]. The author highlights that none of the screening tools currently in use, have incorporated the machine learning algorithm for diagnosing ASD from the recent studies. Along with the problem of unbalanced data, the overlap of ASD, ADHD and Asperger Syndrome as well as different forms of ASD, represents another obstacle in diagnosing, mentioned in [11]. Most studies simplified the classification process by removing these classes and leaving just the ‘Severe Autism’ and ‘Nonspectrum’ as a possible outcome. This of course leads to unreliable classifiers with questionable sensitivity, specificity and accuracy. 3. Problem formulation For the purpose of this project, in total three datasets have been downloaded from UCI Machine Learning Repository [12]. They deal with data related to ASD screening of three different sets of population: toddlers, adolescents and adults. The data was collected through an application in a form of a quiz [13]. The data sets consist of twenty questions, out of which ten are behavioral features, while the other ten are individual characteristics. The application offers four modules representing the age category for individuals from the age: 12-36 months, 4-11 years, 12-16 years and 17 years and older. Ten questions, that differed depending on the age of the individual, from the application are provided in Table 1, in Appendix 1. Description of the data set is provided in Table 2. Table 2. Data set characteristics ASD Screening Data MODULE CHILDREN ADOLESCENTS ADULTS Number of instances 292 104 704 Number of attributes 21 Missing values Yes The first module is based on current parent-assisted ASD screening tool, the Quantitative Checklist for Autism Toddlers (Q-CHAT), while the remaining three are based on appropriate versions of Autism Spectrum Quotient (AQ), which are considered to be good candidates of diagnosing and were somewhat referred to as a ‘red flag’. These screening tools were discussed in [4], where it has been noted that the ten questions can only be used for acknowledging if additional clinical testing is required and is not a definite diagnosis. An analysis that studied these tools is [14]. Next step of this paper will be to investigate the dataset and determine the best way to make the most of it. �4. Machine learning methods Seven algorithms that were chosen for the process of attribute selection with their brief description are provided in this chapter. Bayes Net’s function is to learn the Bayesian networks. This algorithm assumes nominal attributes and no missing values. Search process is done using K2 or TAN algorithm. More sophisticated search methods, used for search, are built on genetic algorithm, hill-climbing, simulated annealing, etc. Search speed can additionally be enhanced by ADTrees [15]. Simple Logistic is an algorithm that builds regression models and fits them with use of LogitBoost and simple regression functions as base learners. Number of iterations are calculated using cross-validation, supporting attribute selection [15]. Decision Stump’s function is to build one-level decision trees for sets with a categorical or numeric value. Missing values in this algorithm are dealt with by seeing them as a separate value and creating a third branch from the stump [15]. J48 is an algorithm that creates a pruned or unpruned C4.5 decision tree. C4.5 This algorithm produces a classifier in a form of a decision tree, which can be either a leaf or a decision node. A leaf indicates a class, a decision node specifies a test with one branch and a subtree for every possible outcome of the test [16]. Logistic Model Tree, or LMT, combines two most popular methods of classification: linear logistic regression and tree induction. This algorithm results in not only classification but also in explicit probability estimates of the class. Another advantage of LMT is that it results in a single tree which makes it easier to interpret [17]. Random Forest is an algorithm that combines tree predictors. Each tree is dependent on values of a random vector, which is sampled independently and with same distribution for all trees of the forest. Generalization error of this algorithm depends on strength of individual trees of thee forest and their correlation. As the number of trees grows, the generalizatioon error converges to the limit [18]. REPTree algorithm represents a fast decision tree learner. This algorithm uses information regarding gain or variance and prunes it with reduced-error pruning to build a ecision or regression tree. Values for numeric attributes are only sorted once, which optimized its speed [15]. 5. Results Classification procedure of this research paper was split in two main parts. First part was applying a 10-fold cross-validation to all attributes and all three datasets. Cross-validation of n-folds splits the original dataset into n parts where n-1 parts are used as a train test, while the nth part is used as a test set [19]. Another method used in this part was applying a percentage split of three different values: 50%, 70% and 90%. Percentage split separates the original dataset into train and test according to the chosen percentage. �Both methods were tested using sixteen algorithms, giving in total 64 results for each dataset. The second part of classification involved attribute selection. Algorithms chosen for this part resulted from the first part, since only those that gave 100% accuracy for all three datasets were again used in attribute selection part. A more detailed 10-fold cross-validation results and algorithm performance are presented in Table 3, and the results of percentage-split and algorithm performance is shown in Table 4. Table 3. 10-fold cross-validation accuracy results Method Cross-validation 10 Algorithm Dataset Bayes Child Adolescent Adult BayesNet 100% 100% 100% NaiveBayes 98.9726% 98.0769% 97.017% MultinomialText 51.7123% 60.5769% 73.1534% BayesUpdateable 98.9726% 98.0769% 97.017% Logistic 95.2055% 95.1923% 97.017% MultilayerPerceptron 99,6575% 89.4231% 100% SimpleLogistic 100% 100% 100% SMO 100% 89.4231% 100% 88.3562% 90.3846% 94.8864% DecisionStump 100% 100% 100% HoeffdingTree 100% 99.0385% 99.858% J48 100% 100% 100% LMT 100% 100% 100% RandomForest 100% 100% 100% RandomTree 93.1507% 80.7692% 96.1648% REPTree 100% 100% 100% Functions Lazy IBk Trees �Table 4. Percentage split accuracy Method Percentage Split (50%-50%, 70%-30%, 90%-10%) Algorithm Dataset Bayes Child Adolescent BayesNet 100% 100% 100% NaiveBayes 98.6% 96.5% 96.5% MultinomialText 51.36 % 55.68 % 41.37 % BayesUpdateable 98.6% 96.5% 96.5% Logistic 93.1% 89.7% 93.1% MultilayerPerceptr on 97.2% 97.7% 100% SimpleLogistic 100% 100% SMO 96.5% 100% 98.07 % 61.53 % 98.07 % 100% 100% Adult 100 % 100 % 54.83 % 50% 100% 100 % 100% 98.01 % 74.14 % 98.01 % 100% 100% 98.5% 97% 74.88 % 98.57 % 97.14 % 80% Functions 84.61 % 94.23 % 87.09 % 93.54 % 90% 96.59 % 94.78 % 95.71 % 80% 100% 100% 100% 100% 100% 100% 100 % 100% 100% 100% 95.4% 100% 92.3% 93.54 % 80% 100% 100% 100% 89.04 % 89.7% 86.2% 88.46 % 90.32 % 100 % 95.73 % 94.31 % 94.28 % DecisionStump 100% 100% 100% 100% 100% 100% 100% 100% HoeffdingTree 98.6% 100% 100% 98.07 % 100% 100% 100% 100% J48 100% 100% 100% 100% 100% 100% 100% 100% LMT 100% 100% 100% 100% 100% 100% 100% 100% RandomForest 100% 100% 100% 100% 100% 100% 100% 100% RandomTree 93.8% 94.3% 82.7% 67.3% 74.19 % 100% 90.99 % 100% REPTree 100% 100% 100% 100% 100% 100% 100% 100% Lazy IBk Trees 100 % 100 % 100 % 100 % 100 % 100 % 100 % Algorithms used for the second part of classifying process of this research paper were chosen according to the percentage of accuracy of Table 3. Out of four Bayes algorithms, only BayesNet gave 100%, SimpleLogistic is the only one out of four Function algorithms that proved the best, and lastly, Tree algorithms shown good results with DecisionStump, J48, LMT, RandomForest and REPTree performing in 100% accuracy for all three datasets. These seven algorithms were used in attribute selection part of classification. All three datasets originally had 21 attributes, and the previous two methods mentioned above included all attributes in the process. Attribute selection method [20] is a process of selecting the most relevant attributes and by doing so, reducing the processing time. �In total, five attribute evaluators have been used in attribute selection process. Those were: ClassifierAttributeEval, CorrelationAttributeEval, ReliefAttributeEval, CfsSubsetEval and WrapperSubsetEval. In a combination of these evaluators, three search methods were used: BestFirst, GreedyStepwise and Ranker [15]. ClassifierAttributeEval evaluates the worth of an attribute with use of a user-specified classifier [21]. CorrelationAttributeEval evaluates the worth of an attribute by measuring the correlation between the attribute and the class [21]. ReliefFAttributeEval sampling of instances happens randomly, and the neighboring instances of the same or different class is checked on [15]. CfsSubsetEval evaluates the worth of a subset of attributes by considering the individual predictive ability of each attribute along with the degree of redundancy between them. Missing values can be seen as a separate value or, with proportion to their frequency, its counts can be distributed among other values [15]. WrapperSubsetEval evaluates attribute sets by using a learning scheme. Cross-validation estimates the accuracy of the learning scheme for a set of attributes [15] . ClassifierAttributeEval, CorrelationAttEval and ReliefAttributeEval required Ranker as a search method. In all three datasets, number of attributes chosen for the Ranker was five. CfsSubsetEval and WrapperSubsetEval work using either BestFirst or GreedySetpwise search method. Combining the evaluators with search methods, we obtained 56 results for each dataset. After the attribute selection was performed on the complete set of 21 attributes, all evaluators resulted with 100% accuracy, regardless of the algorithm used. The attributes of all three datasets are presented in Table 5. Table 5. Attributes by number with description Attribute Description 1 - 10 Score of 10 questions 11 Age (number) 12 Gender (male or female) 13 Ethnicity (list provided) 14 Born with jaundice 15 Autism in family 16 Country of residence 17 Used app before 18 Result of app (automated calculation) 19 Age description (toddler, child, adolescent, adult) 20 Relation (who is completing the test) 21 Class ASD/NoASD Attribute that was present in all three datasets and that showed extremely high correlation was the 18th attribute. This attribute represents the score of ten questions of the application [13]. Therefore, a new approach was used. The 18th attribute was removed completely, and the process of selection was repeated for all three datasets. Results of selection are shown in Table 6, 7 and 8, in Appendix 2. Results of accuracy are shown in Table 9, 10 and 11, in Appendix 3. �The lowest performance for child, adolescent and adult dataset was achieved by DecisionStump, resulting in 78.082%, 70.192% and 82.822% respectively, as can be observed from the results. The lowest number of attributes selected is 1, and the highest is 14. However, the best results required less than that. The algorithms that showed best performance for child dataset were SimpleLogistic and LMT. These algorithms, with applied CfsSubsetEvaluator, resulted in accuracy of 98.973%, and used 10 attributes. BayesNet showed best results, with applied ClassifierAttributeEvaluator, it performed in 90.385% accuracy for adolescent dataset and used only 5 attributes. Simple Logistic successfully classified the adult dataset, with impressive accuracy of 99.432% and used 11 attributes in the process. 6. Conclusion The conclusion is split into two parts, one regarding actions taken to review already written papers and discuss their results, and second which deals with actions taken to derive our own conclusion through processing datasets. This research paper involved three datasets: child, adolescent and adult, with each having 21 attributes. Original datasets were processed using two methods for splitting the dataset into train and test and used sixteen algorithms for both. The obtained results from the first test helped choose algorithms for the second part of testing which involved attribute selection. According to the results, seven algorithms stood out. Attribute selection was performed on all three datasets using seven evaluators. All results had 100% accuracy, despite using different number of attributes. This leaded to another approach which included removing the 18th attribute and reapplying the selection process. Number of attributes for best performances were dependent on the dataset and therefore are different. One should keep in mind that classification process using five attributes can only be used as an indicator of whether further medical testing should be conducted. The main lesson learned, reading papers written so far on this topic, is that integrating ML in ASD diagnosis and its screening tools is a much harder process than it seems. Finding the appropriate number of features and managing to reduce the time of diagnosis depends on many parameters. Many experiments, in an attempt to reduce the time required for the diagnosis process, have discarded some important issues for the sake of simplicity. Their initially admiring results could not be taken for granted, due to misbalanced data and questionable reliability. The algorithm should not be dependent on data. The issue of distinguishing between ASD and PDD related disorders (ADHD, Asperger Syndrome) represents a big obstacle for proper diagnosis of ASD. The algorithm should be provided with a similar number of all possible outcomes in order to learn to better distinguish between categories. The need to reevaluate the current diagnosis tools and adjust them to fit the new criteria from DSM-5 arises. �APPENDIX 1 Table 1. Questions from the ASDQuiz application [13] 13-36 months TODDLER 4-11 years – CHILD 12-16 years – ADOLESCENT 17 & older – ADULT 1. Does your child look at you when you call his/her name? He/she often notices small sounds when other do not? He/she notices patterns in things all the time? I often notice small sounds when others do not? 2. How easy is it for you to get eye contact with your child? 3. 4. 5. 6. 7. Does your child point to indicate that he/she wants something (e.g. toy out of reach)? Does your child point to share interest with you? (e.g. pointing at an interesting sight) Does your child pretend? (e.g. care for dolls, talk on a toy phone) Does your child follow where you are looking? If you or someone else in the family is visibly upset, does your child show signs of wanting to comfort you/them? (e.g. gives a hug) He/she usually concentrates more on the whole picture rather than the small details? In a social group, he/she can easily keep track of several different people’s conversation? He/she finds it easy to go back and forth between different activities? He/she usually concentrates more on the whole picture rather than the small details? In a social group, he/she can easily keep track of several different people’s conversation? If there is an interruption, he/she can switch back to what he/she was doing very quickly? I usually concentrate more on the whole picture, rather than the small details? I find it easy to do more than one thing at once? If there is an interruption, I can switch back to what I was doing easily? He/she doesn’t know how to keep a conversation going with his/her peers? He/she frequently finds that he/she doesn’t know how to keep a conversation going? I find it easy to read between the lines when someone is talking to me? He/she is good at social chit-chat? He/she is good at social chit-chat? I know how to tell if someone listening to me is getting bored? When he/she reads a story, he/she finds it hard to work out the character’s intentions or feelings? When he/she was younger, he/she used to enjoy playing games involving pretending with other children? When I’m reading the story, I find it difficult to work out the character’s intentions? Would you describe your child’s first words as: typical, unusual, the child doesn’t speak? When he/she was on preschool, he/she used to enjoy playing games involving pretending with other children? She/he finds it difficult to imagine what it would be like to be someone else? I like to collect information about categories of things (e.g. types of car, types of bird, types of train, types of plants, etc.) 9. Does your child use simple gestures (e.g. waves goodbye)? He/she finds it difficult to work out what someone is thinking or feeling just by looking at their face? He/she finds social situations easy? I find it easy to work out what someone is thinking just by looking at their face 10. Does your child stare at nothing with no apparent purpose? He/she finds it hard to make new friends? He/she finds it hard to make new friends? I find it difficult to work out people’s intentions? 8. �APPENDIX 2 Table 6. Attribute selection results - child dataset Algorithm s vs. selected attributes BayesNet SimpleLo gistic DecisionS tump J48 ClassifierAttrib uteEval CorrelationAttrib uteEval 4,9,8,10,1 4,9,10,8,6 ReliefAttribu teEval 4,1,10,8,9 CfsSubset Eval (BF & Greedy) 1-10 LMT RandomF orest REPTree WrapperSubs etEval (BF Wrappe r, Greedy 1-10,17 3,4,6,7,10,12, 16 4,10 4,6,10,1 2,16 4 4 1,3,4,5,7,8,10 4,10 ,14 4,6,10,12,16 15,7,8,9,10,15, 4,10 17 1,4,8,10 4,10 Table 7. Attribute selection results - adolescent dataset Algorithm s vs. selected attributes ClassifierAttrib uteEval CorrelationAttrib uteEval ReliefAttribu teEval CfsSubset Eval (BF & Greedy) BayesNet SimpleLo gistic DecisionS tump J48 LMT RandomF orest REPTree 5,4,3,10,6 WrapperSubs etEval (BF Greedy 1,2,3,4,5,7,10 ,12 1,3,4,5,1 0,12 3,4,5,10,14 5,4,10,3,6 5,4,10,3,6 5,3,10,4,8 3,4,5,6,7,8, 10,17 5 2,5,9,14 3,4,5,10,14 5,4,3,10,6 15,7,8,10,17 5,4,10,3,6 2,3,5,8 4,5,7,10 Table 8 Attribute selection results - adult dataset Algorithm s vs. selected attributes BayesNet SimpleLog istic DecisionSt ump J48 LMT RandomFo rest REPTree ClassifierAttribu teEval CorrelationAttrib uteEval ReliefAttribut eEval CfsSubset Eval (BF & Greedy) 9,6,16,19,7 WrapperSubs etEval (BF Greed y 1-10,15,16 1,3,5,6,9,12,1 1,3,5, 5 6,9 9 9,6,8,7,19 9,6,5,4,3 5,9,6,4,7 1-10,16 1-5,710,16,15 1,3,5,6,9,12,1 5 9 1,5,9 1,3,5, 6,9 1-10,12,15,18,19 1,2,3,5,6,7,9,15 �APPENDIX 3 Table 9. Attribute selection accuracy - child dataset Attribute evaluator ClassifierAttribu teEval CorrelationAttE val ReliefAttributeE val CfsSubsetEval WrapperSubsetE val (BF) WrapperSubsetE val (Greedy) Bayes Net 86.644 % 84.932 % 85.616 % 95.206 % 91.438 % 82.192 % SimpleLog istic DecisionSt ump J48 LMT 78.082% 85.95 9% 84.93 2% 83.90 4% 91.43 8% 85.27 4% 82.19 2% 85.95 9% 84.58 9% 87.32 9% 98.97 3% 95.89 0% 83.90 4% 86.301% 84.589% 88.014% 98.973% 95.890% 82.192% RandomF orest 85.274% 81.849% 85.274% 93.151% 89.384% 84.247% REPT ree 85.616 % 85.959 % 83.219 % 83.562 % 84.589 % 83.562 % Table 10. Attribute selection accuracy - adolescent dataset Attribute evaluator ClassifierAttribu teEval CorrelationAttE val ReliefAttributeE val Bayes Net CfsSubsetEval 87.5% WrapperSubsetE val (BF) WrapperSubsetE val (Greedy) 88.462 % 77.885 % 90.385 % 89.423 % 82.692 % SimpleLog istic DecisionSt ump 89.423% 70.192% J48 80.76 9% 81.731% 89.423% 86.539% 68.269% 78.846% 84.61 5% 72.11 5% 71.15 4% LMT 89.42 3% 81.73 1% 88.46 2% 85.57 7% 78.84 6% RandomF orest REPT ree 86.539% 75% 85.577% 83.654% 86.539% 78.846% 76.923 % 68.269 % 72.115 % 71.154 % Table 11. Attribute selection accuracy - adult dataset Attribute evaluator ClassifierAttribu teEval CorrelationAttE val ReliefAttributeE val CfsSubsetEval WrapperSubsetE val (BF) WrapperSubsetE val (Greedy) Bayes Net 90.341 % 91.051 % 90.767 % 96.307 % 95.313 % 92.046 % SimpleLog istic DecisionSt ump 89.347% 90.909% 84.517% 90.625% 99.432% 95.881% 82.822% 93.04% J48 LMT 88.21 % 89.35 7% 90.90 9% 90.62 5% 99.00 6% 95.02 8% 93.89 2% 89.06 3% 92.04 6% 89.34 7% 89.06 3% RandomF orest REPT ree 89.347% 87.5% 91.761% 90.057% 91.193% 94.46% 91.761% 90.199 % 89.205 % 86.364 % 89.063 % 89.921 % �7. References [1] National Institute of Mental Health, Available at: https://www.nimh.nih.gov/health/topics/autism- spectrum-disorders-asd/index.shtml [2] Lord, C., Rutter, M., & Le Couteur, A. 1994. Autism Diagnostic Interview-Revised: A revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. Journal of Autism and Developmental Disorders, 24, 659–685 [3] Lord, C, Risi, S, Lambrecht, L et al, 2000. The Autism Diagnostic Observation Schedule-Generic: a standard measure of social and communication deficits associated with the spectrum of autism J Autism Dev Disord. 2000; 30:205–223.7. [4] F. Thabtah, 2017. Autism Spectrum Disorder Screening: Machine Learning Approach and DSM-5 Fulfillment, ICMHI '17, May 20-22, 2017, Taichung City, Taiwan © 2017 Association for Computing Machinery. ACM ISBN 978-1-4503-5224-6/17/05 [5] American Psychiatric Association 2000. Diagnostic and statistical manual for mental disorder (4th edn), Text revision. Washington, DC: American Psychiatric Association [6] American Psychiatric Association. 2013. Diagnostic and statistical manual of mental disorders: DSM-5. Washington, D.C: American Psychiatric Association [7] C. Lord et al., 1994. Autism Diagnostic Interview-Revised: A Revised Version of a Diagnostic Interview for Caregivers of Individuals with Possible Pervasive Developmental Disorders, Journal of Autism and Developmental Disorders, Vol. 24, No. 5, 1994 [8] L. D. Wiggins, J. Baio, C. Rice, 2006, Examination of the Time Between First Evaluation and First Autism Spectrum Diagnosis in a Population-based Sample, Developmental and Behavioral Pediatrics, Vol. 27, No.2, April 2006. [9] Wall DP, Dally R, Luyster R, Jung J-Y, DeLuca TF (2012) Use of Artificial Intelligence to Shorten the Behavioral Diagnosis of Autism. PLoS ONE 7(8): e43855. doi: 10.1371/journal.pone.0043855 [10] D. Bone, M. S. Goodwin, M. P. Black, et al, 2014. Applying Machine Learning to Facilitate Autism Diagnostics: Pifalls and Promises, Springer Science + Business Media New York 2014 [11] F. Thabtah (2018): Machine learning in autistic spectrum disorder behavioral research: A review and ways forward, Informatics for Health and Social Care, DOI: 10.1080/17538157.2017.1399132 [12] UCI Machine Learning Repository, Available at: https://archive.ics.uci.edu/ml/index.php [13] Thabtah, F. (2017). ASDTests. A mobile app for ASD screening. www.asdtests.com [14] C.Allison, B. Auyeung, S. Baron-Cohen, 2012, Toward Brief “Red Flags” for Autism Screening: The Short Autism Spectrum Quotient and the Short Quantitative Checklist in 1000 Cases and 3000 Controls, Journal of the American Academy of Child & Adolescent Psychiatry, DOI: 10.1016/j.jaac.2011.11.003 [15] I.H. Witten, E. Frank (2005), Data Mining: Practical Machine Learning Tools and Techniques, Elsevier [16] J. Ross Quinlan, C4.5: Programs for Machine Learning, Avilable at: https://link.springer.com/article/10.1007/BF00993309 [17] University of Waikato, Logistic at:https://www.cs.waikato.ac.nz/ml/publications/2005/LMT.pdf Model Trees, Available �[18] L. Breiman (2001), Random Forests, Available at: https://www.stat.berkeley.edu/~breiman/randomforest2001.pdf [19] Towards Data Science, Cross-Validation in Machine Learning, K-Fold Cross Validation, Available at: https://towardsdatascience.com/cross-validation-in-machine-learning-72924a69872f [20] Machine Learning Mastery, An Introduction to Feature Selection, Available https://machinelearningmastery.com/an-introduction-to-feature-selection/ [21] Weka – Version 3.8.3, Waikato Environment for Knowledge Analysis, ClassifierAttributeEval at: � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Machine Learning in Autism Spectrum Disorder Diagnosis Author Author Naida Nalo, Jasmin Kevrić Abstract A summary of the resource. This paper represents an overview of Machine Learning techniques used in Autism Spectrum Disorder - ASD diagnosis. ASD is detected based on behavioral screening which is time consuming and can only be taken by a medical professional. The idea is to find a smaller number of features that are still able to equally well provide satisfying results and not lose the accuracy, sensitivity nor specificity. Some of the algorithms mostly used in recent studies were Artificial Neural Network - ANN and Alternating Decision Trees - ADTrees. The researches usually use WEKA software package for applying the algorithm and obtaining results. Keywords Keywords. Machine Learning, Autism Spectrum Disorder, diagnosis, features, ANN, ADTree, WEKA. Identifier An unambiguous reference to the resource within a given context 2637-2835 Publisher An entity responsible for making the resource available International Burch University, Sarajevo, Bosnia and Herzegovina Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering Date A point or period of time associated with an event in the lifecycle of the resource January, 2020 https://eprints.ibu.edu.ba/files/original/c4bae5fbec77718447ec73dc65649ad6.pdf 62737558fce1a681d66b8438c8f8fedb PDF Text Text Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020214 Handwriting digit recognition using Decision Tree Classifiers Demir Korać1, Samed Jukić1, Mujo Hadžimehanović1 1 International Burch University, Sarajevo, Bosnia and Herzegovina demir.korac@stu.ibu.edu.ba samed.jukic@ibu.edu.ba mujo.hadzimehanovic@stu.ibu.edu.ba Abstract – The usage of handwritten character recognition has been useful for usage from large to common consumer usage. The transitional period of the handwritten to the digital age can be largely improved by focusing on perfecting handwritten character recognition. This paper and work aims to focus on handwritten digit recognition using the decision tree classifier machine learning method, implemented, trained and tested on the data set gathered from the Modified National Institute of Standards and Technology dataset. The data to be recognized is inputted from a pre-existing reliable set, used both for training and testing, in order to give a fair result. The system is run through a Python script and the data set is stored in CSV format, preprocessed and ready for further usage. Taking into consideration the size of the dataset (42000 rows of data), the system’s overall performance is satisfactory with an accuracy of 85% and outputs the results in an understandable manner. Keywords – character, decision, handwritten, recognition 1. Introduction Optical character recognition is the process of mechanically or electronically converting typed, printed or handwritten images of text into a machine-encoded text. The source can be taken from a photo of a document, a scanned document or a scene photo (billboards in a landscape photo). Another possible source is superimposed text on an image, such as subtitle text from a television broadcast. This process of recognition and the technique used in it is called a handwriting recognition system. In literature this recognition is classified into offline handwriting recognition and online handwriting recognition depending on the style of recognition. If an image of handwriting is previously acquired and recognized after it will be classified as offline recognition. However if the handwriting is inputted directly into the machine to be recognized this is called online recognition. One way of doing this is writing on a touchpad or other device dedicated for writing and having it recognized. Another classification exists regarding the recognition and it is based on the process of recognition itself and the technique. We can recognize two main categories: segmentation free and segmentation based recognition. Segmentation free recognition is based on recognizing the character without segmentation into smaller units or characters, �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020214 i.e. words into characters. Segmentation based recognition is the opposite of this. In this approach each word is segmented into smaller units/characters and each character is recognized separately. Handwriting has been for the most part of our history, the primary means of communication and information organization, but with the modernization of these fields handwriting is becoming slowly obsolete. However the legacy of handwritten information cannot be overlooked, so a transitional period and process needs to occur. That’s where handwriting and optical character recognition comes into play. The importance of handwriting recognition is reflected in many industries such as: health care (7,000 people are killed per year by the poor handwriting of doctors) automotive (digital handwriting solutions allow drivers to write characters or numerals, or simply gesture with their fingertips on vehicles' onboard computer screens instead of typing on a standard keyboard) field services (field service technicians use HWR technology for digital data capture, decreasing paperwork and information loss while also allowing technicians to see precise notes on customer history) education (using HWR technology, students can benefit from more than just the increased comprehension linked to taking handwritten notes. For example, handwriting recognition tech can take your sloppy algebra equation, convert it into neat, digital text, and then crunch the numbers in a matter of seconds. ) consumer (with the rapid success of tablets and smartphones, the market is desperately in need of an alternative to inaccurate, digital keyboards and HWR tech is the best remedy.) 2. Previous work Perwej Y. and Chaturvedi A. have worked on recognizing the English alphabet handwriting using Artificial Neural Networks using binary pixels of the alphabet to train the Neural Network and the accuracy of this method is 82.5% [1]. The data set that they worked on are handwritten English alphabet characters which are scanned from documents and then “cleaned” and “smoothed”. The characters are then split into 25 segment grids, scaling and thinning the segments of the characters to obtain skeletal patterns. This is then transformed into binary values representing the segments input. (Shown in Figure 1.) �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020214 Input Alphabet Figure 1. Process of recognition Amrouch M. and Y. Es saady have worked on a method using sliding window based on the Hough transform as a feature extraction technique [2]. An inputted image is divided into two windows and then a dominant direction is determined based on the Hough transform and a directional feature vector sequence has been formed. This method is based on continuous HMMs and directional features with an average accuracy of 90.4%.The data set that they worked on is a database of Amazigh printed characters, containing 240 isolated characters. The images of characters are in Gray level of dimensions 96x96 pixels. M. Hanmandlu and O.V. Ramana Murthy [3] have done a study on recognition of Hindi and English numerals. The numerals are represented in the form of exponential membership functions which serve as a fuzzy model. They have achieved an overall recognition rate of 95% for Hindi numerals and 98.4% for English numerals. The data set that they worked on is a database of 5000 samples of numerals for handwritten English numerals. For Hindi they used a database of totally unconstrained handwritten numerals created using the services of a large number of writers, since there is no standard database available at the moment for handwritten Hindi numerals. Nafiz Arica at al. has proposed a method of recognition without pre-processing which he believed leads to loss of necessary information [4]. This has been backed up with a powerful segmentation algorithm with utilization of character boundaries, maxima and minima, slant angle, upper and lower baselines, stroke height and width and ascenders and descenders which improved the search algorithm of the optimal segmentation path, applied on a gray-scale image. The dataset used was the handwritten database of LancesterOslo/Bergen, which contains single author cursive handwriting. In this dataset, 1,000 words with lower-case letters are segmented and used for HMM training and another disjoint set of 2,000 words are used for testing performance of the proposed system. Table 1. Results of recognition for the LOB Dataset TEST DATA SIZE LOB Dataset 2000 LEXICON SIZE 50 1000 30000 40000 92.3 90.8 89.1 88.8 The overall recognition rate of the whole system on word basis for various lexicon sizes is shown in the table above. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020214 3. Method and Materials The dataset being used in the project is the famous Modified National Institute of Standards and Technology database of handwritten digits [5], as it is the most reliable and largest database of this type. This dataset is taken as a subset of another larger dataset by NIST. The MNIST database of handwritten digits consists of 42000 rows of data. It will be split into a 80/20 ratio for training and test data respectively, so the training data will have 33600 rows and test data will have 8400 rows of data. The digits have been size-normalized and centered in a fixed-size image. The database and the files such as the training set images and labels, and the test set images and labels can be found at: http://yann.lecun.com/exdb/mnist/ The machine learning model used will be a Decision Tree Classifier implemented through the scikit-learn machine learning library in Python [6]. Decision Trees are used in data mining, machine learning and statistics as a predictive modeling approach. The structure of trees is as follows: Class labels are represented as leaves and the conjunctions of features that lead the class labels are represented as branches. The process of operation will be such that the dataset will be inputted as a matrix containing cell inputs from the database as intensity values of pixels of 28x28 images. Table 2. Sample from dataset label pixel0 pixel1 pixel2 pixel3 pixel4 pixel5 pixel6 pixel7 pixel8 pixel9 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 0 0 0 0 0 9 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 An empty classifier will be created and using the fit method it will be filled with the training data. After the classifier finishes with the training data, we will move on to the rest of the data set, the testing part. Using the predict method we will output the classifier prediction of the handwritten digit from the test dataset along with the actual image of the digit. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020214 4. Process The process of recognition is as follows: The dataset is imported and separated into two parts with a ratio of 80/20. The train set consists of 33600 and the test set consists of 8400 rows of data. Splitting the dataset: traindata=data[1:33600,1:] train_label=data[1:33600,0] The data and its label are fitted to the model to learn from it. We do this by using the fit method and passing our training set to it. Fitting the training set: clf.fit(traindata,train_label) The user of the system gives an input of the order number he wants to test the system with, from the dataset. That number is taken and a sample with that order number is chosen from the dataset, along with its actual label. Figure 2. Taking input from the user Using the predict method the sample is predicted from the sample and the prediction is outputted. Using the shape method we are taking the sample from a row vector and shaping it as a 28x28 matrix of pixel intensity values. We are then creating a figure, which will be displayed after prediction. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020214 Using the shape method: d.shape=(28,28) When the figure is created, we are proceeding with the prediction. We will then be using the predict method from the Decision Tree Classifier in the scikit-learn library [7]. Using the predict method along with output of the data: print ("The predicted digit is =", clf.predict( [testdata[number]])) print ("The actual digit is =", actnum) Along with predicting the digit, we will also be outputting the actual label of the sample number that the user chose. An extra feature added is the current certainty of prediction calculation. Certainty of prediction: p=clf.predict(testdata) count=0 for i in range (0,8400): count+=1 if p[i]==actual_label[i] else 0 print ("Certainty of prediction=", (count/8400)*100,"%") For this feature we are taking all numbers in the range of the test data set and using the predict method, as to calculate the success rate of our model on the current test data. Figure 3. The systems’ process from the users’ perspective �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020214 5. Usage workflow In Figure 4. We can see the workflow according to which the system will be operating. Figure 4. Usage workflow Since the system does not need dataset input from the user, we will just be importing the dataset. The system needs user input, since the user will be choosing the digit that is to be recognized, so he needs to provide the order number from the data. The system then outputs its prediction along with other data such as the actual label and the prediction certainty. This allows the user to analyze the data given to him, and check if the prediction was correct and is the prediction certainty satisfactory. 4. Decision Tree Classifier in scikit-learn A decision tree is a flowchart-like tree structure where an internal node represents a feature (or attribute), the branch represents a decision rule, and each leaf node represents the outcome. The topmost node in a decision tree is known as the root node. It learns to partition on the basis of the attribute value. It partitions the tree in a recursive manner called recursive partitioning. This flowchart-like structure helps you in decision making. It's visualized like a flowchart diagram which easily mimics the human level thinking. That is why decision trees are easy to understand and interpret. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020214 Figure 5. Decision tree operation diagram 5. Results After processing and thorough testing the model seems to have a success rate of ~85% on the test dataset given to it. The system itself may have some confusion when analyzing similar digits. This occurs because the digits themselves are handwritten and because the data has been gathered from over 250 different writers, and some of them write different digits in a similar manner. An example is shown in Figure 6. where we can see that the system is having trouble predicting number 9, but instead predicts 3. From the image we can see that the digits themselves are quite similar. Figure 6. Wrong prediction of digit 9 as digit 3 �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020214 Different parts of the world write digits differently, so the algorithm would be much more accurate if it would be implemented based on location, and also, if different Machine Learning methods or Ensemble methods would be used, we would surely improve accuracy. Finally the thing that would take the recognition to the next level is image quality, considering that this algorithm has been implemented on a dataset with quite low quality of images. 5. Conclusion It should be noted that this data set is not sufficient for a higher level model, to be used in production, but as its using a small dataset, used for initial training, it has shown to be a pretty successful model and method. If we would train the aforementioned algorithm on a larger dataset with higher image quality, it could be used for real world projects. The confusion of the system may also be affected by the quality of the images in the given dataset, as it is limited to only 28x28 matrices. Future training with a dataset with matrices of larger magnitude may prove more successful. REFERENCES [1] Perwej Y. & Chaturvedi A. -Neural Networks for Handwritten English Alphabet Recognition. 2011 IJCA [2] M. Amrouch, A. Rachidi, M. El Yassa, D. Mammass - Printed amazigh character recognition by a hybrid approach based on Hidden Markov Models and the Hough transform. 2009 IEEE https://scikitlearn.org/stable/modules/generated/sklearn.tree.DecisionTreeClassifier.html [3] M. Hanmandlu, O.V. Ramana Murthy - Fuzzy model based recognition of handwritten numerals. 2007 Pattern Recognition [4] N. Arica, F.T. Yarman-Vural - Optical character recognition for cursive handwriting. 2002 IEEE [5] MNIST handwritten digit database - http://yann.lecun.com/exdb/mnist/ [6] Decision Tree Classification in Python - https://www.datacamp.com/community/tutorials/decision-tree-classification-python [7] sklearn.tree.DecisionTreeClassifier- � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Handwriting digit recognition using Decision Tree Classifiers Author Author Demir Korać, Samed Jukić, Mujo Hadžimehanović Abstract A summary of the resource. The usage of handwritten character recognition has been useful for usage from large to common consumer usage. The transitional period of the handwritten to the digital age can be largely improved by focusing on perfecting handwritten character recognition. This paper and work aims to focus on handwritten digit recognition using the decision tree classifier machine learning method, implemented, trained and tested on the data set gathered from the Modified National Institute of Standards and Technology dataset. The data to be recognized is inputted from a pre-existing reliable set, used both for training and testing, in order to give a fair result. The system is run through a Python script and the data set is stored in CSV format, preprocessed and ready for further usage. Taking into consideration the size of the dataset (42000 rows of data), the system’s overall performance is satisfactory with an accuracy of 85% and outputs the results in an understandable manner. Keywords Keywords. character, decision, handwritten, recognition Identifier An unambiguous reference to the resource within a given context ISSN 2637-2835 Publisher An entity responsible for making the resource available International Burch University, Sarajevo, Bosnia and Herzegovina Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering Date A point or period of time associated with an event in the lifecycle of the resource January, 2020 https://eprints.ibu.edu.ba/files/original/b07ea9e57cba6ae337bfd4515469085d.pdf 1aead1a15145473c159b69bbe91031a6 PDF Text Text Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020213 Biometrics Based Access Control System Mujo Hadžimehanović1, Dino Kečo1, Demir Korać1 1 International Burch University, Sarajevo, Bosnia and Herzegovina mujo.hadzimehanovic@stu.ibu.edu.ba dino.keco@ibu.edu.ba demir.korac@stu.ibu.edu.ba Abstract – Access control includes attendance checking and intrusion prevention. It is used to protect property, employees and other assets of a company or institution. Since attendance checking and intrusion detection are important segments of many educational institutions and other businesses as well, it is important to make these processes faster, easier and as convenient as possible. Lots of institutions are suffering from unreliable attendance checking methods, so we have decided to use biometrics, more precisely face recognition to automate and improve this overall process. As part of this study the full system has been implemented for recognition of people. As an example of usage in an educational institutions multiple photos will be recorded during the class session, so that in case of students leave class after the first shot, they will be removed from the attendance sheet. All recognized people will be stored in Mongo database as an array of features and later read from database and processed by using Python script for face recognition. All educational institutions are going to have benefits from this study. Benefits would be improving attendance management and security. Keywords - attendance, face, images, recognition 1. Introduction Technology is rapidly improving nowadays and everyday activities are adopting these improvements. Point is to automate these activities and not to lose time performing them. Attendance is a really important part in most organizations such as schools, faculties, companies etc. Today, attendance is performed in various ways. Best way to do it is biometrics. Biometrics is a bioengineering area which is an automated method for person recognition based on its physiological or behavioural characteristics. There are many biometric templates such as fingerprints, face, hand geometry, iris, voice or signature. System is going to use face biometric template because it is the fastest approach and requires no human intervention. This method is better than other biometric methods because these methods are time consuming. There are also lots of systems which are using RFID cards, location based attendance tracking systems, signature based etc. Negative things about these methods are that they can be faked. In RFID and location based systems employees are carrying RFID cards or GPS locators. So, other people can check instead of other employees. There are two main stages of face recognition process and they are face detection and face identification [1]. Recording employees' work hours and their activities, attendance of students in schools are really important components of every company or school. This process is maintained by using signature, fingerprint, iris, RFID or face recognition. System is going to use a camera which captures images of people entering the company or school building. Detected faces will be compared with pictures which are already in the database. If a person's picture is in the database attendance �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020213 will be checked automatically. Otherwise, the security system is going to be informed that an unrecognized person entered the building. So, this system can also be used as an intrusion detection system. 2. Literature Review Following section contains a presentation of all related work and their methods. Since they have large-scale data with massive noisy labels, X. Wu, R. He, Z. Sun, and T. Tan used a Light CNN framework [2]. Their Light CNN architecture contains Max-Feature-Map to suppress low activation neurons in all layers. Their model was trained on Celeb-1M dataset. In order to handle noisy labeled images, they proposed a semantic bootstrapping method to automatically re-label training data via pre-trained deep networks. For training purposes they used five types of databases. First type is commonly used Labeled Faces in the Wild which consists of ~13,250 images of ~5,750 people. At VR@FAR=0 for Light CNN-29, they achieved 97.50%, while results from all other methods were lower than 70%. Next type of the database are collections of images extracted from Youtube videos which contain YouTube Celebrities (YTC) and Celebrity1000 database. Precision achieved for these datasets is 94.18%. Third type are MegaFace, IJB-A and IJB-B datasets which are challenging and they got 85.13% precision. Cross-domain databases are the fourth type of database. It includes CACD - VS, Multi - PIE and the CASIA NIR - VIS 2.0 database. They achieved 98.55% on CACD, 95.0% on Multi-PIE and on CASIA VR@FAR=0.1% result is further improved from 94.03% to 94.77%. M. Arsenovic, S. Sladojevic, A. Anderla, and D. Stefanovic use Convolutional Neural Networks (CNNs) cascade to detect faces and CNN to generate embeddings of each face [3]. Fact is the best results for larger datasets are achieved by using CNNs, but in their production environment that was not the case. CNN gave the best results for smaller datasets. Accuracy of 95.02% was achieved on a dataset created by authors in the realtime environment. Five employees of the company took pictures of themselves and they used these pictures as a dataset. Model was trained with these 5 pictures. Active annotation and learning framework was used by H. Ye et al [4]. They are starting with face image training set without labels and train a deep neural network iteratively model created was used to choose examples for further manual annotation. After following active learning strategy, Value of Information criterion is derived to actively select candidate annotation images. This model reaches the coverage of 70.7% with a precision of 95%. MSR Image Recognition Challenge by J. Li et al introduces a knowledge base which has an idea to assign each face unique entity key and provide large dataset consisting of about 100,000 famous persons with around 100 images per person (MsCeleb) [5]. Method achieved coverage of 46.1% at 95% precision on the random dataset and 33% at 95% precision on the hard set of their challenge. Authors proposed a method consisting of two stages to learn robust human face representations for effective recognition of human faces. First stage in the training set is cleaning the noisy data because dataset is taken from the internet so images without faces can appear. In order to do so, a deep neural network was trained on existing dataset. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020213 S. Chintalapati and M. V. Raghunadh used SVM and Bayesian classifier for automated attendance system based on face detection and recognition [6]. They proved that these classifiers are better when compared to other distance classifiers. This system automatically detects the student which enters the classroom and marks the attendance if recognizes him. One of the failures of the system is recognizing faces only up to 30 degrees angle variations. 3. Methods and Materials Dataset to be used is Labeled Faces in the Wild [7] which is a database of face photographs designed for studying the problem of unconstrained face recognition. It contains more than 13,000 pictures collected from the web. Each image has been labeled with the name of the person on it. There are 1680 different persons in images. Fig.1 shows samples from LFW dataset. Fig.1. Samples from dataset These images need to be processed in order to get numerical representation of faces which is called feature vector. Feature vector consists of various numbers in a specific order which can be: height or width of face, width of lips, nose height etc… Final output of processed image needs to be an array with features which is shown in Fig.2. All features are stored in the mongo database for speed improvement. Python script iterates in a folder which has dataset images and stores one by one in a database with image name and features. Face Recognition library with deep learning is going to be used for this project. Deep learning model has an accuracy of 99.38% on Labeled Faces in Wild dataset. Features of face recognition library are finding faces in pictures, finding facial features in pictures and identifying faces in pictures. Once installed face recognition gives us two command line programs: ● face_recognition - recognize face on image ● face_detection - detect face in image Face recognition process consists of two stages. This includes taking and preparing training dataset and integration into existing system. For testing purposes, data was collected at the university. These are images of students which were taken in the first year. Images are preprocessed and inserted into the mongo database by �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020213 using python script for inserting images. After insertion, facebook images of the same people were taken and tested by using python script for face recognition. Fig.2. Array of features 3.1 Data preprocessing Implemented system is going to use monitoring cameras at the entrance. It means that we could have some kind of network or other problems and taken images could be blurry, so we have to include such images in training dataset, Fig.3. Persons entering the building can be photographed from different angles, so these kind of photos should be included also in training dataset, Fig 4. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020213 Fig.3. Original and Blurred image Fig.3 shows an example of original and blurred images. If we have perfect conditions we would have a picture like the original one, but if the system experiences network issues we might have a blurry image like the one represented. System has to be ready to respond accordingly to these kinds of issues. Python script was written using OpenCV [8] interface to generate blurry images out of the original ones. Fig.4. Image from front and side In Fig.5 we can see facial features drawn on picture of Pep Guardiola. Most important features are shown: eyes, nose, mouth and chin location. These features are used when recognizing people on images. Face recognition library contains script face_landmarks for detecting facial landmarks and positioning the face based on them. Fig.5. Facial features �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020213 3.2 Usage Workflow Application usage workflow is represented on Fig.6. Images need to be collected into a single folder, so that insertion helper scripts can be run. For multiple insertions we need to pass a folder of images to script, which iterates through these images, creates encodings and inserts them into database. Single insertion script accepts an image as a parameter, encodes it and inserts into a database. Last step is recognizing images, the script accepts an image which needs to be recognized and iterates through mongoDB collection and looks for matching images. Fig.6. Usage workflow 3.3 Face Recognition Library Face recognition library which we are using is built using dlib’s face recognition with deep learning. We can install library by using a python package installer. Once installed we are provided with two command-line programs : face_recognition and face_detection. Face recognition recognizes faces in a photograph or in a folder of photos. There should be two folders, one containing known people and second which contains photos of people which we want to recognize. Face recognition program is run with two parameters which are the names of these folders. Face detection program finds pixel coordinates of faces. It takes a folder with images as parameter and at the end prints one line for each face that was detected. There is also an option to speed up the overall process by doing a recognition with multiple CPU cores. For example if we have 8 core CPU, we can process 8 times as many images in the same amount of time. Dlib is a toolkit written in C++ and contains ML algorithms and tools for solving real world problems. The most important thing is that dlib is an open source library which enables anyone to use it anywhere, free of any charge. Some of the dlib’s features are Deep Learning, Multiclass SVM, Image Processing etc.. Our library uses Image Processing tool for face recognition built by using deep learning tools from dlib. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020213 3.4 Helper Scripts There are multiple helper scripts which enable user to insert multiple or single image into database and recognize faces. The most important parts of scripts are represented in the following lines. insertMultiple.py for image in images: current_image = face_recognition.load_image_file("images/" + image) encodings = face_recognition.face_encodings(current_image) if len(encodings) > 0: current_image_encoded = encodings[0] num_of_images+=1 else: print("No faces found in the image " + image) num_of_not_found+=1 continue mydict = { "image": image, "encoding": current_image_encoded.tolist() } x = mycol.insert_one(mydict) print(image + " inserted") Code snippet above loads images from the folder, encodes them and inserts them into mongoDB. We have to provide the name of the folder which contains images and simply run the script. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020213 faceRecognition.py unknown_image = face_recognition.load_image_file("image.jpg") unknown_face_encoding = face_recognition.face_encodings(unknown_image)[0] known_faces = [] names = [] for x in mycol.find(): known_faces.append(x['encoding']) names.append(x['image']) results = face_recognition.compare_faces(known_faces, unknown_face_encoding) for x in range(len(known_faces)): if results[x] == True: print("Recognized: " + names[x]) else: print("Failed: " + names[x]) Code represented above does face recognition. It takes an image of the person which we want to recognize, encodes it, loops through face encodings from the database and checks if a person exists in the database. 4. Results By using a custom dataset which was collected from our university. Students' images were taken and tested on created scripts. From these tests we have obtained accuracy of 90.9% when testing on images found on Facebook. There were some problems when recognizing people from different angles, but this can be material for further study. Images of people are not shown because they did not agree to publish their images. Since face_recognition python library has a pre-trained model there is no need for additional training. Improvement is that all images are inserted in mongoDB with image name and face encodings array. Fig.5 shows one part of mongoDB record. Python script for inserting images in mongoDB is written and it takes a folder with images and inserts one by one in the database. In our testing environment LFW dataset is used and all images are collected into a single folder. Number of images inserted in the database is 4014. There are also images on which faces are not recognized. Unrecognized images number is 21. So, if we take into consideration that 4014 images are inserted and 21 are unrecognized which means that more than 99% images were recognized. Example of such an image is shown in Fig.6. Execution time of the script is about 35 minutes for LFW Dataset. Final result of this research would be access control application. Application can be installed on Raspberry Pi which has a camera installed. All assets that are necessary for access control application to be fully functional can be installed on Raspberry. These are mongoDB, python and python libraries. Overall process is not so challenging, so that we do not need anything better than Raspberry Pi 3 B+ which is a model that we used while testing the application. �Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020213 5. Conclusion The aim of this study was to make the attendance checking process a lot easier for companies and schools by using biometrics. Every employee or student would be recorded by camera at the entrance and recorded in the system. For educational institutions cameras will be installed in classrooms so that the system can make multiple shots during lessons. This research was successful because it made the recognition process faster by using a document based database which is really fast. This study will bring benefits for multiple groups. Benefit for schools is easier attendance recording and reducing waste of time at the beginning of the classes. Also students will not have a chance to avoid coming to classes because this system will not allow them to cheat. Similar benefit is for companies to track their employees coming and leaving time. Future research suggestions in this field are solving problems if a person is recorded from the side and possibly getting blurry images because of internet connection issues. REFERENCES [1] B. T. Liyew and P. Hazari, “A Survey on Face Recognition based Students Attendance System.” [2] X. Wu, R. He, Z. Sun, and T. Tan, “A Light CNN for Deep Face Representation With Noisy Labels,” IEEE Transactions on Information Forensics and Security, vol. 13, no. 11. pp. 2884–2896, 2018, doi: 10.1109/tifs.2018.2833032. [3] M. Arsenovic, S. Sladojevic, A. Anderla, and D. Stefanovic, “FaceTime — Deep learning based face recognition attendance system,” 2017 IEEE 15th International Symposium on Intelligent Systems and Informatics (SISY). 2017, doi: 10.1109/sisy.2017.8080587. [4] H. Ye et al., “Face Recognition via Active Annotation and Learning,” Proceedings of the 2016 ACM on Multimedia Conference - MM ’16. 2016, doi: 10.1145/2964284.2984059. [5] J. Li et al., “Robust Face Recognition with Deep Multi-View Representation Learning,” Proceedings of the 2016 ACM on Multimedia Conference - MM ’16. 2016, doi: 10.1145/2964284.2984061. [6] S. Chintalapati and M. V. Raghunadh, “Automated attendance management system based on face recognition algorithms,” 2013 IEEE International Conference on Computational Intelligence and Computing Research. 2013, doi: 10.1109/iccic.2013.6724266. [7] “LFW Face Database : Main.” [Online]. Available: http://vis-www.cs.umass.edu/lfw/. [Accessed: 19-Jan-2019]. [8] “OpenCV library.” [Online]. Available: https://opencv.org/. [Accessed: 09-Jan-2019]. � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Biometrics Based Access Control System Author Author Mujo Hadžimehanović, ​Dino Kečo,​ ​Demir Korać​ Abstract A summary of the resource. ​Access control includes attendance checking and intrusion prevention. It is used to protect property, employees and other assets of a company or institution. Since attendance checking and intrusion detection are important segments of many educational institutions and other businesses as well, it is important to make these processes faster, easier and as convenient as possible. Lots of institutions are suffering from unreliable attendance checking methods, so we have decided to use biometrics, more precisely face recognition to automate and improve this overall process. As part of this study the full system has been implemented for recognition of people. As an example of usage in an educational institutions multiple photos will be recorded during the class session, so that in case of students leave class after the first shot, they will be removed from attendance sheet. All recognized people will be stored in Mongo database as an array of features and later read from database and processed by using Python script for face recognition. All educational institutions are going to have benefits from this study. Benefits would be improving attendance management and security. Keywords Keywords. Keywords - attendance, face, images, recognition Identifier An unambiguous reference to the resource within a given context 2637-2835 Publisher An entity responsible for making the resource available International Burch University, Sarajevo, Bosnia and Herzegovina Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering Date A point or period of time associated with an event in the lifecycle of the resource January, 2020 https://eprints.ibu.edu.ba/files/original/a7be93f106bd4655fce2c143192ca565.pdf 671002f6cf69f8a0bc016d089e362b6e Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Transient Stability Analysis on Modified IEEE 14-Bus System Author Author Ermin Šunj, Ammar Arpadžić, Mirza Šarić Abstract A summary of the resource. Power system stability represents an important condition of the safe and efficient operation of the electric power system. This paper presents the transient stability analysis for the case of the new generator grid connection. The analysis is performed on a modified IEEE 14-Bus test system. In total, two cases were analysed. In the first case, the analysis with the maximum installed power of the generator (250 MW) is carried out, while in the second case, the analysis is performed using the optimal generator ratings (75 MW) using the DIgSILENT PowerFactory software. The transient stability analysis was carried out under the three-phase symmetrical faults and the N-1 criterion requirements. The results indicate that power system large disturbances significantly influence system operation and characteristics. This paper demonstrates the importance of transient stability analysis, which is an important part of power system studies and must be included in generator grid connection approval. Keywords Keywords. active power, IEE 14-Bus system, N-1 criterion, reactive power, rotor angle, transient stability Identifier An unambiguous reference to the resource within a given context 2637-2835 Publisher An entity responsible for making the resource available International Burch University, Sarajevo, Bosnia and Herzegovina Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering Date A point or period of time associated with an event in the lifecycle of the resource January, 2020 https://eprints.ibu.edu.ba/files/original/cda47db38a0256d84f45336ef7e1fa5e.pdf 9d17336775d02b9918c005099df3eb63 PDF Text Text Journal of Natural Sciences and Engineering, Vol. 2, (2020) DOI number: 10.14706/JONSAE2020211 Impact of Electric Vehicles in a Grid-to-Vehicle Mode on Voltage Stability Naida Nalo1, Emina Kišija1, Mirza Šarić1 1 International Burch University, Sarajevo, Bosnia and Herzegovina naida.nalo@stu.ibu.edu.ba emina.kisija@stu.ibu.edu.ba mirza.saric@ibu.edu.ba Abstract — With a rapid development and a massive deployment of electric vehicles, the power system is facing many challenges regarding power quality and voltage stability. This paper deals with the impact of electric vehicle in grid-to-vehicle mode depending on different EV penetration levels and point of connection on static voltage stability impact of a real low-voltage distribution network. Based on nine variations created, results showed that connecting vehicles closer to the beginning of the feeders creates a smaller voltage drop, therefore more vehicles can be connected. However, going farther from the feeder causes voltage to go below 0.9 p.u. and eventually leads to instability. Keywords—electric vehicles, grid-to vehicle, static voltage stability. 1. Introduction Power system stability can be defined as the ability of the system to remain in an equilibrium state under normal operating conditions and to regain that equilibrium state after being subjected to a physical disturbance [1]. According to [2] power system stability is defined as a term applied to alternating – current electric power systems, denoting a condition in which the various synchronous machines of the system remain in synchronism, or „in step“, with each other. On the opposite side, instability is defined as a condition involving loss of synchronism or falling „out of step” [2]. However, instability can also occur without the loss of synchronism [1]. Electric vehicles (EVs) are considered to be a promising solution both for reducing air pollution and also as being introduced as a new form of distributed generation when working in a vehicle-to-grid (V2G) mode. Many countries are offering incentives and by doing so, motivating EV owners to charge their vehicles in scheduled times, to help flatten the daily peaks [3]. In the stability calculations, the behavior of the system under the effect of a transient disorder is of interest. Equipment reacts as a system response to a disorder. In each situation, only part of the protection reacts, therefore, the problem must be simplified and the key factors for each type of instability must be defined. �In this project, the problem of voltage stability is investigated in a LV distribution network involving 46 buses and 42 loads. According to different penetration levels of (EVs) and point of connection across the network, nine scenarios were modeled and examined for voltage instability. 1.1 Voltage stability Voltage stability is the ability of the system to maintain acceptable voltage values on all busbars in the system, both in normal operating conditions and after the effects of the disruption. Voltage instability occurs when a disorder, which can be caused by an increase in consumer demand or a change in operating conditions, causes a progressive and uncontrollable voltage drop. The main cause of voltage instability is the inability of the system to respond to reactive power requirements. The core of the problem is usually a decrease in the voltage in the flow of active and reactive power through inductive reactances representing the transmission network [1]. The criterion for voltage stability is that on all busbars in the system, under certain operating conditions, the bus voltage is increasing as injection of reactive power on the same buses is increasing. Therefore, the system is unstable if voltage level is decreasing as reactive power is increasing, at least on one busbar in the system. Voltage stability is a local phenomenon, but its consequences may have a widespread impact. A voltage collapse is far more complex than voltage instability and according to [4], can be explained as an inability of the power system to supply the reactive power or as an excessive absorption of reactive power by the system itself. It can also be defined as a process in which voltage instability causes very low voltage levels in a substantial part of the system. A local voltage collapse can and will lead to a widespread collapse of the power system [4]. 2. Literature review Electric vehicles have experienced a warm welcome by pollution-aware society. Their non-polluting nature helped them gain popularity and become one of the most sold cars in Norway, according to the Norwegian Road Federation [5]. However, deploying large fleets of electric vehicles impacts the load profile of the network since EVs are introduced as additional loads when being connected for charging [3]. A study in [6] focused on the static voltage stability impact of EV charging stations. A cluster load model equivalent to 20 sets of EV chargers was taken into consideration along with the different probability distributions of state of charge (SOC) of the batteries. The authors concluded that charging stations are most likely to cause voltage instability due to the variableness of power during the charging process. Research conducted in [7] focused on the static voltage stability of plug-in EVs with respect to different charging methods. The results showed that voltage stability is closely related to the proportion of the constant impedance and the constant power load. Since EVs were considered as constant power load, the less the initial voltage drop percentage, the more EVs will be allowed to access the distribution network. �Authors of [8] investigated the power quality and dynamic stability aspects of vehicle to grid connection of EVs which uses a bidirectional power flow and allows the EVs to give back to the grid if needed. Their conclusion was that charging and discharging state of the PEV does not affect negatively neither the voltage stability nor frequency since they remain within allowed limits. A study in [9] investigated the impact of high PV penetration in a low-voltage distribution network on voltage stability. In the paper, PV curves were used to analyze the static voltage stability in a test node of an important and possibly critical line. It was shown that the node situated near the end of the network had the weakest PV characteristics due to power loading and the distance from the feeder. However, they concluded that integrating photovoltaic units with 40% penetration level would optimize the voltage stability of that node. Research conducted in [10] analyzes voltage stability with aid of PV curves on an example of a real transmission network. EVs included in the study were all charged during the daily peak load with a six-hour charge time. Results showed that high levels of EV penetration, with the expected annual increase, leads to unacceptably large voltage variations. A new method for analyzing the impact of PEVs in distribution networks was proposed in [11]. As in many other papers, this study confirmed that a small number of vehicles does not create stability issues whereas a large fleet of vehicles causes a greater effect on the grid. Charging strategies as the overall conclusion was highlighted in this paper. A study in [12] investigated an impact of EV charging on voltage variations and unbalance in a real low voltage distribution network. Different scenarios were created to depict several EV penetration levels and load distributions across the network. Results confirmed the work of other papers, showing that point of connection plays an important role to the level of impact of EVs to two analyzed power quality parameters. An analysis, similar to [9], will be conducted in this paper, using PV curves to determine the critical busbars along the two feeders of the low-voltage distribution network. Section III explains the modeled network and created variations, Section IV draws results and Section V draws conclusions. 3. Methods 3.1 Problem formulation Electric vehicles in this project are treated as single-phase loads connected to the network, in a grid-to-vehicle (G2V) mode. Because of the increasing number of EV charging stations being integrated to the power system, analysis of their clustering effect and influence on the static voltage stability have become important and necessary. In this project, analysis of an impact of EV charging on voltage stability is performed on a real example of a part of a distribution network. �3.2 P-V Curves In voltage stability studies, characteristics of interest are the relationships between transmitted power P, receiving end voltage V and reactive power injection Q. P-V and Q-V curves are traditional forms of displaying these relationships. In this project, P-V curves are analyzed. Power-Voltage analysis process includes increasing transfers of power (MW) and monitoring what happens with voltages in the system. This is done by increasing the power system load and, at each increment, power flows are recomputed (P-V curve is non-linear and full power flow solutions are required) until the nose of the PV curve is reached, that is, the maximum transferred power [13]. That point represents the critical voltage because after that, rapid decline of voltage occurs. Therefore, reaching maximum power is highly avoided because operating at or near stability limit risks a large – scale blackout. Power margin between the current operating point and critical voltage is used as voltage stability criterion [14]. 3.3 LV Distribution Network Modeling and Variations In this paper, the analysis was done using the model of 46 – bus LV distribution network, with total of 42 loads distributed along two feeders, modeled in DIgSILENT Power Factory. Modeled network is provided in Appendix 1. Length of the first feeder is 371 m while the length of the second feeder is 253 m. Nine variations were modeled, including: low, medium and high EV penetration levels, at the beginning, in the middle and at the end of the network. List of variations is provided in Table 1. Numbers of EVs included in each variation are presented in Table 2 and Table 3. Table 1. Network variations (penetration-point of connection) Case no. Network Variations 1.1 Low-beginning 1.2 Medium-beginning 1.3 High-beginning 2.1 Low-middle 2.2 Medium-middle 2.3 High-middle 3.1 Low-end 3.2 Medium-end 3.3 High-end 1 2 3 Table 2. Number of EVs in Variations Network Variations Variati on No. 1.1 2.1 3.1 1.2 2.2 3.2 Penetration Level Number of EVs Percentage of penetration level Low 6 ≈15% Medium 12 ≈30% �Table 3. Number of EVs in Variations cont’d Network Variations Variati on No. 1.3 2.3 2.4 Penetration Level Number of EVs Percentage of penetration level High 21 50% 4. Results 4.1 Case 1 - EVs distributed at the beginning First three variations were modeled so that electric vehicles are placed near the beginning of the two feeders. Each variation had a different penetration level of EVs as explained in Table 1. After the load flow calculation was performed, Transmission Network Toolbox was activated, and PV curves were calculated. To see which busbars stay within the allowed limits and which do not, several busbars were selected from the beginning, middle and end of each of two feeders and included in the resulting PV graph. The obtained graphs for Case 1 variations are presented in Figure 1, Figure 2 and Figure 3, respectively. Fig. 1. PV curves for Variation 1.1 �Fig. 2. PV curve for Variation 1.2 Fig. 3. PV curve for Variation 1.3 All busbars whose PV curves are above the voltage value of 0.9 p.u. are acceptable and stable, while those below 0.9 p.u. are not stable and therefore not acceptable. As presented in the graphs, two busbars, plotted in blue and grey, have values below 0.9, which makes them unstable. These two busbars are from the first feeder, situated in the middle and at the very end of the feeder. All busbars from the second feeder stayed within allowed limits, as well as the busbar from the beginning of the first feeder. �4.2 Case 2 – EVs distributed in the middle Three variations examined for the impact of EV charging and placement around the middle of the two feeders were 2.1, 2.2 and 2.3 Number of EVs connected to the feeders are with respect to Table 1. Several busbars were selected and included in resulting PV graph, to depict the voltage stability across the two feeders, that is, to show how stable are busbars from the beginning, middle and end of the two feeders. Results for the abovementioned variations are shown in Figure 4, Figure 5 and Figure 6. Fig. 4. PV curve for Variation 2.1 �Fig. 5. PV curve for Variation 2.2 Fig. 6. PV curve for Variation 2.3 According to results obtained from the PV graph, conclusions similar to those in previous variation can be drawn. All busbars from the second feeder and only the busbar from the very beginning of the first feeder stay within allowed limits of stability, that is above 0.9 p.u. value of voltage, shown in the y-axis. Two busbars from the middle and at the end of the first feeder show instability. �4.3 Case 3 – EVs distributed at the end Last three variations from Figure 2 were modeled to investigate how much EVs connected near the end of the feeders will affect voltage instability of selected busbars across the two feeders. Number of connected vehicles per variation is shown in Table 1. Selected busbars remained the same as those used in the previous six variations. Results obtained are shown in Figure 7, Figure 8 and Figure 9. Fig. 7. PV curve for Variation 3.1 �Fig. 8. PV curve for Variation 3.2 Fig. 9. PV curve for Variation 3.3 Results from the obtained PV graph of the last tested variations show that voltages of the two terminals from the middle and end of the first feeder experience a drop below 0.9 p.u. but gets a more constant value when compared to results of previous variations. All busbars from the second feeder and only one from the very beginning of the first feeder have values greater than 0.9 p.u., making them well within allowed limits of voltage stability. �5. Conclusion The purpose of this paper was to analyze the impact of different EV penetration and points of connection on voltage stability of a distribution network. Load flow analysis was performed on all nine scenarios followed by a PV curve calculation in Transmission Network Toolbox of DIgSILENT. Then, a static voltage stability analysis was performed using PV curves for a number of selected busbars from the beginning, middle and end of the two feeders. The criterium was that all curves above 0.9 p.u. value of voltage were acceptable, and all below show voltage instability. It was found that the length of the feeder, point of connection and level of EV penetration played a crucial part when it comes to voltage stability. All selected busbars from the second feeder remained within allowed limits of voltage values while only one busbar from the beginning of the first feeder was above the limiting value. This was mainly due to the length of the feeder, amount of loading and the distance from the source feeder. For large fleets of EVs being connected and charged at the same time, voltage stability and power quality becomes of crucial importance. Distribution system operators must pay attention to the impacts of charging on power quality and stability of the distribution system, especially if vehicles are in close range, and situated farther from source. If no modifications are to be made to increase the network’s capacity, then only a limited number of vehiclescan be allowed, with reference to the point of connection. A possible solution could be integration of small distributed generators or implementation of renewables, dispersed along the network to decrease voltage variations and increase power quality, especially near the end of the feeders, where critical nodes are. Future work might include analyzing and modeling the impact of connection of photovoltaics or small wind generators in terms of distributed generation, which are expected to improve the voltage levels and overall variations. �APPENDIX 1 �6. References [1] N. Rajaković, Analiza elektroenergetskih sistema II, Elektrotehnički fakultet, 2007 [2] E.W. Kimbark, Power system stability (Vol. 1). John Wiley & Sons, 1995 [3] J.Y. Yong, V.K. Ramachandaramurthy, K.M. Tan, N. Mithulananthan, A Review on the State-Of-Art Technologies of Electric Vehicle, Its Impacts and Prospects, Elsevier, Renewable and Sustainable Energy Review 49 (365-385) 2015 [4] P. Kessel, H. Glavitsch, Estimating the Voltage Stability of A Power System. IEEE Transactions on Power Delivery, 1(3), 346-354, 1986 [5] Electric vehicles are now the majority of cars sold in Norway, (20198, April 1st), Retrieved on 27th of May, from: https://www.autoblog.com/2019/04/01/electric-vehicles-are-now-the-majority-of-cars-sold-in-norway/ [6] M. Zhang, J.H. Zheng, W.Z. Wang, M.T. Dai, Research on Static Voltage Stability Based on EV Charging Station Load Modeling, the international conference on Advanced Power System automation and Protection, 2011 IEEE [7] Y. Zhang, S. Xiaohui et al., Research of Voltage Stability Analysis Method in Distribution Power System with Plug-In Electric Vehicle, PES Asia-Pacific Power and Energy Conference, 2016 IEEE [8] E. Alghsoon, A. Harb, M. Hamdan, M., Power Quality and Stability Impacts of Vehicle To Grid (V2G) Connection, The 8th International Renewable Energy Congress (IREC 2017) [9] M. Ghaffarianfar, A. Hajizadeh, Voltage Stability of Low-Voltage Distribution Grid with High Penetration of Photovoltaic Power Units, Energies 2018, 11(8), 1960, Available at: https://doi.org/10.3390/en11081960 [10] S. Avdaković, A. Bosović, Impact of Charging a Large Number of Electric Vehicles on the Power System Voltage Stability, Elektrotehniki Vestnik,137-142, 2014. [11] Y. Kongjeen, K. Bhumkittipich, Impact of Plug-in Electric Vehicles Integrated into Power Distribution System Based on Voltage-Dependent Power Flow Analysis, Energies 2018. [12] N. Nalo, A. Bosović, M. Musić, Impact of Electric Vehicle Charging on Voltage Profiles and Unbalance on Low Voltage, Advanced Technologies, Systems, and Applications IV – Proceedings of the International Symposium on Innovative and Interdisciplinary Applications of Advanced Technologies (IAT 2019), vol 83. Springer [13] C. Reis, F.M. Barbosa, A Comparison of Voltage Stability Indices. MELECON Mediterranean Electrotechnical Conference (pp. 1007-1010). IEEE, May 2006 [14] C.A. Cazares, Voltage Stability Assessment: Concepts, Practices and Tools. IEEE/PES Power System Stability Subcommittee Special Publication 2002 (SP101PSS). � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Impact of Electric Vehicles in a Grid-to-Vehicle Mode on Voltage Stability Author Author Naida Nalo, Emina Kišija, Mirza Šarić Abstract A summary of the resource. With a rapid development and a massive deployment of electric vehicles, the power system is facing many challenges regarding power quality and voltage stability. This paper deals with the impact of electric vehicle in grid-to-vehicle mode depending on different EV penetration levels and point of connection on static voltage stability impact of a real low-voltage distribution network. Based on nine variations created, results showed that connecting vehicles closer to the beginning of the feeders creates a smaller voltage drop, therefore more vehicles can be connected. However, going farther from the feeder causes voltage to go below 0.9 p.u. and eventually leads to instability. Keywords Keywords. electric vehicles, grid-to vehicle, static voltage stability Identifier An unambiguous reference to the resource within a given context 2637-2835 Publisher An entity responsible for making the resource available International Burch University, Sarajevo, Bosnia and Herzegovina Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering Date A point or period of time associated with an event in the lifecycle of the resource January, 2020 https://eprints.ibu.edu.ba/files/original/b35517f5a6b3b1b5430cbd3beb65665b.docx 3d88bd71483956752150c8fc8f575fa4 Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource THE RETURN OF THE MODERN Author Author LUKA KORLAET SVEBOR ANDREJEVIC Abstract A summary of the resource. The second half of nineties and twenties were the building Eldorado: the bank loans were shared without any delays to entrepreneurs and apartment buyers. It was built a lot and very uncertainly. Speaking of Zagreb, the slopes of Medvenice were especial ly attractive. That space was built with numerous urban vil las, the smal l multi-storey buildings, which its investors and architects were led by exploation logic and rarely think about quality of floor plans or about environment between buildings. Publisher An entity responsible for making the resource available International Burch University Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering https://eprints.ibu.edu.ba/files/original/d98c62dc674b330134eb6d0d5eec8c89.docx b7b81d52bcb066ce1b2d555ea001830b Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal of Natural Sciences and Engineering Identifier An unambiguous reference to the resource within a given context 2637-2835 DOI Digital object identifier 10.14706 Publisher An entity responsible for making the resource available International Burch University Description An account of the resource Journal of Natural Sciences and Engineering (JONSAE) is a peer-reviewed, biannually published international journal focusing on empirical and theoretical research in all branches of Engineering and Natural Sciences. It is published on the behalf of Faculty of Engineering and Natural Sciences of International Burch University and aims to provide the best content regarding by publishing original research papers, review articles, special issues, feature articles, and book reviews. All manuscript submissions are subject to initial appraisal by the Editor, and, if found suitable for further consideration, to peer review by independent, anonymous referees. All peer review is double-blind and submission is online. The journal welcomes theoretical, applied, interdisciplinary and methodological work, with preference on empirical research, critical approach and problem-solving methods in manuscripts. Language A language of the resource English Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Communication in Smart Homes with Emphasis wn Power Line Communication Author Author Ina Salihović Esma Musić Dejan Jokić Abstract A summary of the resource. Abstract: Power Line Communication (PLC) is a technology that allows consumers to use the already existing wiring infrastructure to exchange information. This paper overviews narrowband PLC in home automation, starting from the basics of power line communication and its advantages compared to wired and Wi-Fi automation systems, data modulation techniques, noise problems, frequency bands, all the way to regulations affecting PLC. The paper is finished off with an overview of three System on Chip (SoC) power line modems from a few different generations, Yitran’s IT800D from 2005, ON Semiconductor’s NCN49597 from 2012, and STMicroelectronics’s ST8500 from 2017. Keywords Keywords. Keywords: Power Line Communication, narrowband PLC, PLC modem, System on Chip Publisher An entity responsible for making the resource available International Burch University Source A related resource from which the described resource is derived Journal of Natural Sciences and Engineering