Alican Bozyiğit

Abstract: Recommendation systems guide users to choose the most appropriate items among numerous alternatives based on predicting their interests. Recently, it is seen that recommendation systems have become to be widely used in educational domain, especially in course recommender applications. The objectives of these systems is facilitating course selection process of students and reducing their stresses. The current course recommendation studies generally consider the most recent grades of the courses taken by students and ignore the case of repeating the course under the pass-fail or grade replacement options. However, retaking a course is the primary parameter giving opinion about tendency of the students to the courses. In this study, we propose a novel collaborative filtering (CF) based course recommendation system considering the case of repeating a course and students' grades in the course for each repetition. We experiment different Ordered Weighted Averaging (OWA) operators which aggregates grades for each student's repeated courses to enhance the recommendation quality. The normalized mean absolute error (MAE) of our approach using CF and OWA is calculated as 0,063 which is encouraging for future work.

Abstract: Public transport route recommendation is a complex problem because passengers take many factors into consideration while planning their trips. In this paper, a novel route recommending approach is proposed to find the ideal public transport route with respect to multiple factors such that number of transfers, total distance, and walking distance (in the order of importance). Space P modelling technique and a Dijkstra's Algorithm based method are applied together for the first time by this approach. The proposed method is tested on the real-world dataset (Public Transport Network of Izmir, Turkey) having 7,704 stations and 43,467 connections between these stations. In the experimental results, it is clearly seen that our method finds the optimal route regarding the specified factors for each given destination in milliseconds.

Abstract: The popularity of social media applications supporting activities such as communication and sharing contents is increasing rapidly. Although social media facilitates sharing information between people, it can be used for malicious purposes. Cyberbullying, such as flaming, harassment or denigration, is one of the most important types of social media misuse. In this study, it is aimed to detect Turkish cyberbullying messages on social media. In this direction, a compressive dataset is created and published on the internet, since there is no publicly available dataset for Turkish cyberbullying contents. Then, correction of the misspelled slang words, feature extraction and selection, and popular machine learning techniques for text mining are applied respectively. It is observed that Naïve Bayes Multinomial, support vector machines, and k nearest neighbor are most successful techniques for detection of Turkish cyberbullying contents. The results of the study are quite successful and motivating for future work.

Abstract: In this study, we determined five factors that we thought to be directly affecting a course selection and investigated the level of importance of these factors according to both undergraduate and graduate students, and experts (i.e. academicians). We conducted two separate questionnaires for students and experts to this end. Then, by applying Analytic Hierarchy Process (AHP) we obtained three separate sets of weights for the determined factors. Each set of weights represents the preferences of a group of participants i.e. undergraduate students, graduate students or experts. This study helps understanding the needs and opinions of the user groups of course recommender systems. In future, we intend to develop a course recommender system using the determined factors and their respective weights.

Abstract: Public transport is preferred by most of the people since it provides various advantages. As a result, many route-planning applications are developed for users of public transport. The general aim of these applications is proposing the ideal route for a given destination; however, there are various route-planning criteria for public transport. According to our research, "the number of transfers" is seen as the primary criterion for the route planning by most users. In this study, an approach for public transport is proposed in order to recommend the route that is comprised of minimum number of transfers. In this approach, Space P and a pareto optimal solution are used for modelling network, then Breadth First Search is modified to plan the ideal route on the modelled network. Furthermore, the proposed approach is experimented on the public transport network of Izmir. It is proved that from any source to any target, our route recommender returns the path with the minimum number of transfers optimally within milliseconds.

Abstract: Public transport applications, which aim to propose the ideal route to end users, have commonly been used by passengers. However, the ideal route for public transport varies depending on the preferences of users. The shortest path is preferred by most users as a primary criterion for the ideal route. According to our research, Dijkstra's Algorithm is mostly used in order to find shortest path. However, Dijkstra's Algorithm is not efficient for public transport route planning, because it ignores number of transfers and walking distances. Thus, in order to minimize these shortcomings, Dijkstra's Algorithm is modified by implementing penalty system in our study. Additionally, our modified algorithm is tested on the real world transport network of Izmir and compared with the results of Dijkstra's Algorithm. It is observed that our modified algorithm is quite efficient for route planning in the public transport network in terms of the number of transfers, distance of proposed route and walking distance.

Abstract: In this study, an evaluation methodology for route planning applications in public transportation is given, and existing web applications that aim to make the people use public transportation effectively, are evaluated in terms of running speed, ease of use, effectiveness etc. While inspecting outputs of software systems, Analytic Hierarchy Process (AHP) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is used to obtain accurate and consistent comparison results. In conclusion we obtained important factors for users; e.g. running time of finding solution and accuracy of the route returned as the solution have a major share of total 38% between factors. Also we observed general deficits of the applications in factors; e.g. accuracy of the route returned as the solution has 66% success rate.