Theses supervised by Dr. Öğr. Üyesi Murat Saran
10 theses · Çankaya University
Identifying management errors that lead to failure in it projects and recommendations for project managers
Information technology (IT) projects bring significant benefits to organizations if they are successful, but some IT projects face setbacks before or after delivery and fail. IT projects can fail for many reasons. This study reveals that most IT projects fail due to various reasons, including mismanagement of projects by IT project managers, lack of customers' knowledge of technical details, and lack of communication between the project owner and the project team. This study presents a literature review on the causes of failure in IT projects by examining online resources, which are scientifically accepted databases such as ProQuest and EBSCOhost, at the first stage. Afterward, a survey was conducted for people working in the IT sector to reveal the reasons for the failure of IT projects. This survey also tried to determine the management mistakes of IT project managers in their IT projects and other factors that lead them to failure. Finally, it was tried to make suggestions to the project managers to carry out their IT projects with minimum failure by identifying the issues that caused the failure of the IT projects and the management errors.
Gamification in e-learning: The effect on student performance and perception at an Iraqi university
Creating an effective environment in e learning is one of the challenges that are encountered by educational and pedagogical institutions to increase the engagement and motivation of students during the educational process. One method to make an e learning environment effective and attractive is the application of the gamification concept, which improves learners' engagement by integrating e learning with game design elements. The aim of this study is to analyze the effects of gamification elements such as (points, level-up, badges and leaderboard) in Moodle system on students' performance and perception. Forty-seven Iraq University students were participated in the online Architecture course into two different settings. Thirty students of the experimental group worked with gamification tools (points, level-up, badges and leaderboard) and seventeen students of the control group had access to the same session and activities but without the gamification tools. Data sources included students' grades of pre and post-tests on papers and students' survey results of experimental group according to the Technology Acceptance Model (TAM). The results indicated that students in the experimental group had no statistically significant difference from the control group regarding the student's performances on pre-test scores, whereas the experimental group students had higher grades than the control group students indicating a statistically significant difference regarding the students' performances on post-test scores. Furthermore, nearly all students in the experimental group strongly agreed that using the gamification tools (points, level-up, badges and leaderboard) in Moodle system were engaging and beneficial in education.
Development of a prescription recommendation system using case-based reasoning
In this thesis, a prescription recommendation system was developed based on past prescriptions in order to reduce the workload of physicians and increase the accuracy of written prescriptions. The case-based reasoning method used in this research is among the technological developments used in real life in different fields. In this study, the prescription recommendation system was developed using case-based reasoning method and research was performed to find out the performance of this system. In order to create a set of data to be used in the study, 7120 anonymous prescription information was collected from 300 volunteers through a website. The success rate of the system was then calculated by comparing the prescriptions (1) with the prescriptions prescribed by ten physicians in different branches, and (2) with the latest prescriptions in the data set by using the nearest neighbors' algorithm. The success rate of the system obtained by comparing the prescriptions of the real-life prescriptions for the patients and the recommended prescriptions is 0.78. Besides, the success rate of the system regarding the comparison of the last prescriptions of the 50 most common diseases in the system's data set and the prescriptions recommended by the system is 0.91. The results of this study indicate that health-care professionals can benefit from the recommendation system developed in this study. In general, with the recommendation system designed in this study, health-care professionals are supported to make faster and more accurate decisions during the prescription writing process.
Scientometric mapping of Türkiye's computer engineering graduate research landscape
This study examines 12,778 master's and PhD theses in Computer Engineering from universities in Türkiye. The theses were completed between 1984 and 2024. The main goal is to show how postgraduate research has grown, changed, and improved over the last 40 years. All theses I used came from the YÖK National Thesis Center, since that is where Türkiye keeps its graduate work. I first cleaned the texts myself with simple steps like joining title+abstract, doing basic lemmatization, and catching a few n-grams. After that, I tried several modelling tools in no fixed order. BERTopic was the main one because it uses neural embeddings, but I also checked LDA and even TF-IDF just to compare what they highlight. They did not produce the same structure, yet together they helped me understand the overall shape of the data. In the end, the models showed nearly ninety topic groups. The number of theses also rose a lot after 2015 as more graduate programs opened. Some themes suddenly became popular too — for example cybersecurity, blockchain, agricultural AI, and medical image work. There was also a clear difference between degree levels. PhD theses usually went toward theory related topics like optimization, privacy, or advanced modelling, while master's work stayed closer to practical tasks. Master's theses are usually more practical and focus on things such as face recognition, mobile apps, or smart home systems. The study also found that 33.6% of the theses do not match the main topic groups very well. These studies can be seen as "unconventional" or outliers. Overall, this study is one of the first large-scale studies of computer engineering research in Türkiye. The results show both the growth in the number of theses and the increasing variety of topics.
Federated learning for credit card fraud detection: A privacy-preserving approach with controlled noise integration
With the rapid increase in e-payment technology, cards are present as one of the essential tools. However, with this growth comes a risk of fraudulent attacks that may cause involved parties' losses and damage. Banks aim to establish strong fraud detection systems to protect assets and obey regulator rules. Therefore, developing a model that copes with security and integrity requirements is crucial. This research project introduced advanced Machine Learning (ML) models, such as Deep Neural Network (DNN) and ensemble learning with AdaBoost, to detect fraud while addressing the skewed nature of datasets using balancing methods. Additionally, it facilitates collaborative learning among banks using Federated Learning (FL) while preserving data privacy. In this study, the FL model was tested against various percentages of label-flip attacks to evaluate resilience against malicious acts by banks trying to sabotage the learning process. The models were tested on two datasets: a real dataset from an Iraqi bank and the known Kaggle creditcard dataset. Models were assessed on a set of performance metrics to cover all aspects of the methods. Results showed that ensemble learning with Random Forest (RF) and AdaBoost achieved remarkable performance across both datasets. Moreover, the FL cosine-based worked better than the existing Federated Average method. Lastly, the proposed approach combining RF+AdaBoost with FL cosine aggregation surpassed the existing method when validated on a private bank dataset achieving 96.47% accuracy and 94.58% recall. The present study contributed to the academic literature by addressing the lack of real datasets in the fraud detection domain.
Identifying research trends in computer engineering and computer science master's programs in Türkiye using topic modeling techniques: LDA, TOP2VEC, AND BERTOPIC
This thesis aims to identify research trends in Computer Engineering and Computer Science Master's programs in Turkey through the application of topic modeling techniques. In this thesis, BERTopic, Top2Vec and LDA methodologies are used to identify the research topics of 6,174 Computer Engineering and Computer Science Master's theses published in Turkey between 2020 and 2024, obtained through the YOK thesis database. According to the results, both LDA and BERTopic techniques yielded the best results in terms of coherence score, while LDA showed superior performance in terms of the perplexity metric. The findings reveal that theses in 2020 mainly focused on data analysis and software applications, while machine learning was a prominent research area in 2021. In 2022, image processing and machine learning topics are prominent, and in 2023, machine learning and algorithm theory. Finally, in 2024, the most popular topics are artificial intelligence and natural language processing. The results of this study provide university administrators with a data-driven methodology and forward-looking insights to align their academic and research agendas with the evolving national landscape of computer engineering and computer science research.
Developing an integrated information security model for civil aviation: A comprehensive framework for risk assessment and mitigation strategies
This thesis proposes an integrated information security model to enhance information security in civil aviation, focusing on the integrity, confidentiality, and availability of computer-based systems in modern civil aviation. The research aims to identify and mitigate potential threats to various critical systems by examining existing and emerging aviation technologies. The study includes an analysis of avionics, flight data networks, mobile systems, and electronic flight bags, as well as ground-based systems that are integral to flight operations. Through a survey of aviation personnel, suggestions on training, awareness, threats, and solutions to information security issues affecting aviation operations were collected, and the information security model needed was tried to be understood. Through this holistic approach, the thesis aims to develop an information security model that addresses the challenges posed by the increasing digitalization of aviation infrastructure. The resulting model is expected to contribute significantly to the national aviation cybersecurity knowledge base and provide valuable insights for industry stakeholders, operators, and researchers. Ultimately, this study aims to enhance civil aviation's overall safety and reliability by strengthening stakeholders' digital awareness against evolving cyber threats.
Application of a voting-based ensemble method for recognizing seven basic emotions in real-time webcam video images
Automatic recognition of human emotions based on facial expressions is a challenging task with significant implications in various fields, including human-computer interaction, healthcare, and affective computing. In recent times, modern deep learning techniques, particularly Convolutional Neural Networks (CNNs), have exhibited promising results in the domain of facial emotion recognition. The study presents comprehensive research to reveal the most effective method for recognizing seven basic emotions from 2D facial images on real-time video or real-time webcam. The research investigates and compares different methods, including Data Augmentation methods, CNN models, KNN models, and a hybrid CNN-KNN approach. The proposed hybrid CNN-KNN method involves leveraging the rich feature representations learned by a pre-trained CNN model for emotion analysis. The pre-trained CNN extracts high-level features from facial images, which are then used as input to a KNN classifier for emotion classification. The thesis evaluates the hybrid CNN-KNN approach against traditional standalone CNN models and KNN models to assess its performance and effectiveness on real-time video. In addition to the hybrid CNN-KNN method, the thesis explores several other approaches to facial emotion recognition, including different CNN architectures, transfer learning using pre-trained models, data augmentation techniques, and ensemble methods. The aim is to thoroughly analyze and compare the performance of these methods and determine the optimal approach for accurate emotion recognition. The assessment is carried out using the FER2013 dataset, a well-established collection of labeled 2D facial images representing seven fundamental emotions. To comprehensively gauge the methods' effectiveness, a range of performance metrics including accuracy, precision, recall, F1 score, real-world experiments and computational efficiency are employed. This research's findings shed light on each approach's strengths and weaknesses and identify the most effective method for facial emotion recognition. The results will guide the development of emotion recognition systems in real-world applications, enabling more empathetic and context-aware human-computer interactions. Our main motivation is investigating the effectiveness of ensemble methods in emotion recognition. Our side motivations are to develop FER systems and start a new DB to be beneficial to both Turkiye and the world, improving communication between humans and machines, making technology more sensitive to human emotion, supporting technology to understand humans and humans to understand technology. As a result of this research, Successfully attained a remarkable 95% accuracy on the amalgamated FER2013, CK+, and KDEF datasets, leveraging a comprehensive support base of 29,716 instances. Introduced a novel database, ATS_FER_DB_2023, achieving a commendable accuracy of 94% on merged new DB and FER2013. This database encompasses 86 images, featuring prominent Turkish celebrities. In the realm of real-time emotion recognition, a meticulous comparison of various methods within our environment revealed that the CNN-KNN algorithms, enhanced with Principal Component Analysis (PCA), emerged as highly effective in our specified system assembly.
Development of case-based recommendation system for course selection in a university context
Nowadays, recommendation systems that guide people for their choices are one of the most widely used software technologies. Especially in the field of education, this technology can be used to make course selection more efficient for students. In this thesis, a recommendation system for the course selection of the students was developed and the accuracy of the recommendations were tested. In the system developed within the scope of this study, Case-Based Reasoning (CBR) method was applied by using the real data of the previous students. While preparing the proposals, the data of students who take similar courses in their curriculum are used. The CBR method presented the suggestions of the courses to be taken by the students and the success rate of the system was calculated by comparing the appropriateness of the relevant suggestions with the lessons actually taken. According to the results of this study, the success rate of the suggestion system is 69%.
The design, development and evaluation of a smart attendance tracking system using bluetooth low energy beacons
The Bluetooth Low Energy (BLE) signal is a wireless personal area network device that periodically broadcasts the signals of a BLE advertising. Smartphone devices take these signals and are used to locate users. This process will allow the provision of context-aware information to users' mobile devices and link the online virtual world to the online physical world. The main purpose of this study is to develop a system that allows students to register their participation in classes using BLE devices and implement this system on a university campus. This research is aimed to reach a set of key targets, including (1) The development of a system capable of polling students by using BLE beacons as an indoor positioning system; (2) Reduce the time for attendance registration; (3) To use more than one BLE device in parallel for students' attendance tracking. This study is also an attempt to understand the potential of using BLE beacons in educational institutions and the challenges in using BLE beacons. In this study, different topologies were designed to evaluate the accuracy of the measured powers from iBeacon devices, and a different number of iBeacon devices were used to do this. Based on measured power levels for iBeacon devices in the class using decision trees and random forest classifiers in each topology, the correctness of estimating the positions of the students were assessed. Power levels were achieved using software developed specifically for Apple iPhone devices; where three measurements were recorded for each location in the class to minimize the error. The results of this study show that students who are positioned close to the classroom walls have the highest probability of failure to register. Moreover, students outside the class can participate in the attendance without being in the classroom. For this reason, these regions are considered to be critical regions, and the signal powers are collected with a half meter resolution to increase the density of the collected measurements in these regions. To provide a useful method, it is essential to distinguish students who are physically located in the class from those who are not; so that the student should not be able to register for participation unless they are actually in the class. For this purpose, closed-field positioning technology based on Bluetooth Low Energy (BLE) devices is used in this study. Different distributions of BLE devices have been evaluated in this study to recommend a topology where students from outside the classroom are never allowed to participate in attendance. In this study, it was found that at least four BLE devices were required for each class to achieve 100% accuracy using random forest classifiers to classify students in and out of the classroom.