Filters

2016 - 201912020 - 20256
Reset filters

Settings

search.filters.applied.f.dc.language.iso: en × Subject: Machine Learning ×

Search Results

Now showing 1 - 7 of 7
  • Thumbnail Image
    ItemEmbargo
    The Art of Personalized Student-Supervisor Matchings
    (ADAL Kitap, 2025) Serek A.G.; Berlikozha B.A.
    The process of student-supervisor matching is a critical yet complex task in higher education institutions, directly influencing research productivity, student satisfaction, and workload distribution. The ability to assign students to the most suitable supervisors is essential for fostering strong academic relationships, optimizing institutional resources, and improving research outcomes. However, traditional manual assignment methods often lead to inefficiencies, subjective biases, and an imbalance in workload distribution. As a result, automated recommendation systems have emerged as a promising solution to enhance the efficiency and fairness of student-supervisor pairings. This study evaluates three recommendation algorithms—Singular Value Decomposition (SVD)-based collaborative filtering, graph-based matching using the Hungarian Algorithm, and machine learning via Random Forest Regression—to determine their effectiveness in optimizing student-supervisor assignments. A rigorous empirical analysis is conducted across five key performance metrics: accuracy, fairness, stability, scalability, and computational efficiency. The findings reveal that while collaborative filtering performs well with established datasets, it struggles with novel cases due to its dependence on prior interactions. The Hungarian Algorithm guarantees optimal matching but faces scalability challenges, particularly in large academic institutions with thousands of students and supervisors. Meanwhile, Random Forest Regression effectively captures complex compatibility patterns but requires extensive labeled data, limiting its applicability in cases where historical matching data is sparse or unavailable. To overcome these limitations, the study proposes an adaptive hybrid framework that integrates the strengths of all three approaches. The hybrid model leverages collaborative filtering’s ability to recognize patterns in existing data, the Hungarian Algorithm’s precision in optimal pairings, and the predictive power of machine learning. By combining these methodologies, the proposed system enhances match accuracy, ensures fair workload distribution, and remains computationally efficient for large-scale institutional implementation. Additionally, the framework introduces dynamic adaptation mechanisms that allow the system to update recommendations based on real-time changes in student preferences and supervisor availability, making it more practical for real-world applications. The research contribution is a comprehensive, empirically validated hybrid framework that improves student-supervisor matching by balancing accuracy, fairness, and efficiency. This study provides educational institutions with actionable guidelines for scalable and equitable assignment processes, ultimately contributing to more effective mentorship experiences, improved research collaborations, and enhanced academic outcomes.
  • Thumbnail Image
    ItemOpen Access
    TRAINING A SINGLE MACHINE LEARNING AGENT USING REINFORCEMENT LEARNING AND IMITATION LEARNING METHODS IN UNITY ENVIRONMENT
    (СДУ хабаршысы - 2020, 2020) Urmanov M. ; Alimanova M.
    Abstract. This paper provides a research of Unity plugin that helps to develop Machine Learning Agents within Unity engine environment. This work introduces training a single Machine Learning Agent using both Reinforcement Learning and Imitation Learning methods, comparing the results and effectiveness.
  • Thumbnail Image
    ItemOpen Access
    GRADE PREDICTING SYSTEM
    (2022 International Young Scholars' Conference, 2022) N. Zhailau
    Abstract Nowadays, prediction of academic performance became necessary for educational entities and universities. As you know most higher educational institutions have a portal system that monitors academic performance. This is necessary in order to assist at-risk students and ensure their retention, as well as to provide exceptional learning resources and experiences, as well as to improve the university’s rating and reputation. Predictive analytics employed advanced analytics, including machine learning implementation, to improve achievement and to generate high-quality performance. As a result, the primary goal of this project is to demonstrate the feasibility of training and modeling on a small dataset size, as well as the feasibility of developing a prediction model with a credible accuracy rate. Using visualization and clustering algorithms, this study investigates the possibility of identifying the key indicators in the small dataset that will be used to create the prediction model.
  • Thumbnail Image
    ItemOpen Access
    ADAPTATION OF GAMING PROCESS TO IMPROVE HAND REHABILITATION
    (Suleyman Demirel University, 2016) Alimanova M.O.; Kozhamzharova D.Kh.; Zholdygarayev A.O.; Meraliyev M.M.
    Hands, as the most dexterous part of our body, are of vital importance to our everyday life. However, since hands are extensively used in nearly all tasks, they are exposed in more dangerous environment than any other parts. Overwork, injury and geratic complications, such as stroke can all cause hand function, totally or partially, which directly diminish the q uality of life. Unpleasant effects caused by trauma and overwork to hands results with immediate hand rehabilitation. Trainings for patients’ rehabilitations are normally goes in rehabilitation center in hospitals, with getting some physiotherapy for han ds, making some exercises and etc. However all this may bore patients and not to motivate to sooner recovery. The Leap Motion controller is a small device that senses consumer gestures and is aimed to enlarge a user’s interactive experience with their computer. Using infrared sensors, it is able to collect data about the position and motions of a user’s hands. This allows to use leap motion in different purposes like development of children’s intellect, having fun with playing virtual reality games and etc. One more example for effective usage of such device is in the purpose of medicine.
  • Thumbnail Image
    ItemOpen Access
    Personalized Career-Path Recommender System for STEM Students
    (Faculty of Engineering and Natural Science, 2024) Zhalgassova Zh.
    This dissertation introduces a Personalized Career-Path Recommender System (PCRS) designed to help high school students in Kazakhstan, particularly those interested in STEM (Science, Technology, Engineering, and Mathematics) fields. The system uses the Myers-Briggs Type Indicator (MBTI) personality types and students’ academic performance to offer personalized recommendations for university specializations. The research addresses the common challenges faced by students, such as high dropout rates and frequent changes in majors, often due to the lack of structured career guidance. To tackle these issues, the study collected a variety of data, including students’ demographics, academic records, and personal attributes, as well as detailed profiles of university majors. Advanced machine learning techniques, including content-based filtering, collaborative filtering, fuzzy logic, and hybrid approaches, were used to process this data and generate accurate recommendations. The effectiveness of the PCRS was tested with real data from students at SDU University. The results show that the system can provide relevant and personalized career guidance, significantly improving students’ decision-making processes and satisfaction with their chosen specializations. By combining MBTI personality assessments with academic performance data, this research offers a fresh approach to educational technology and career counseling. The insights and methods developed in this study can be adapted for use in other regions facing similar challenges, ultimately helping more students make informed and satisfying career choices.
  • Thumbnail Image
    ItemOpen Access
    Mitigating Bias in AI-Based Loan Approval Systems through Fairness-Centric Techniques
    (SDU University, 2025) Raziyeva S.
    As artificial intelligence (AI) becomes increasingly embedded in high-stakes decision-making systems, ensuring fairness in algorithmic outcomes has emerged as a critical concern. This thesis investigates bias and fairness in AI-based credit scoring systems, with a particular focus on gender disparities. Using the German Credit Dataset as a case study, the research evaluates the performance and fairness of several supervised machine learning models, including Logistic Regression, Decision Tree, Random Forest, XGBoost, Support Vector Machine, and Neural Network. The study applies fairness metrics such as Statistical Parity Difference (SPD) and Disparate Impact (DI) to assess group-level inequalities in predicted loan approval outcomes. Results reveal a consistent trade-off between model accuracy and fairness, where high-performing models like Random Forest and XGBoost demonstrate notable biases against female applicants. Even interpretable models, such as Logistic Regression, exhibit fairness issues due to historical and structural biases embedded in the training data. To address these challenges, the thesis highlights the importance of incorporating fairness-aware strategies across the machine learning pipeline, including data pre-processing, fairness evaluation, and potential post-processing mitigation. The use of tools like AIF360 and stratified sampling further strengthens the analysis. This research contributes to the growing discourse on responsible AI by demonstrating that achieving fairness is not merely a technical goal but a sociotechnical imperative. It calls for an interdisciplinary approach that combines ethical reasoning, regulatory compliance, and algorithmic transparency to ensure equitable access to financial services. The findings advocate for the development of AI systems that are not only accurate but also accountable and inclusive.
  • Thumbnail Image
    ItemOpen Access
    Breaking Barriers with AI: The Evolution and Challenges of Automated Sign Language Recognition
    (SDU University, 2025) Joshi M.; Khankriyal P.; Chandola Y.; Uniyal V.
    Communication remains a significant challenge for individuals with hearing impairments and speechrelated disabilities, especially when others are not familiar with sign language. Developing technologies that facilitate seamless communication for these individuals is crucial to promote equality for disabled people and accessibility for all. Sign language recognition systems have emerged as a promising solution, typically implemented using a hardware or software-based approach. Hardware solutions, such as sensor-equipped gloves, often pose usability and cost barriers, making them less appealing for widespread adoption. In contrast, software-driven approaches using artificial intelligence (AI), deep learning (DL) and machine learning (ML) offer a more practical and scalable alternative. This paper provides a complete review of recent developments in AI-based sign language recognition systems, with a particular attention towards deep learning architectures such as Convolution Neural Networks (CNNs). The aim is to evaluate current methodologies, highlight their strengths and limitations, and identify potential directions for future research to improve communication technologies for hearing-impaired people.