Welcome to SDU Institutional Repository

 

Recent Submissions

ItemOpen Access
Evaluation Performance of Students by use Neural Network in Kazakhstan
(Faculty of Engineering and Natural Science, 2013) Halil A.
This thesis evaluates the ability of Artificial Neural Networks (ANN) to predict the performance of Taldykorgan Kazakh Turkish High School students in Kazakhstan. Educational institutions take a big part in people life.School administrators and student's parents pay attention for the performance of the students. Academic researchers have developed several models to predict the improve the performance. Artificial Neural Networks have the potential to provide human characteristics of problem solving that are so difficult to simulate using the analytical. logical techniques of Decision Support System(DSS) . ANN can analyze big quantities of data to establish patterns and characteristics in situations where the rules or logic are not known(Turban,et al.2006). ANN model is designed, built on available student's data. The TANI forecasting program of ANN is used to predict the student's performance based on inputs and output parameters. N This study's result. shows that the neural network model can predict student's real performance with very small errors and predicts better when the today’s and future's samples have similar characteristics
ItemOpen Access
Implementation of Real-time Focus Tracking in Video Streams through Advanced Computational Methods and Image Analysis
(Faculty of Engineering and Natural Science, 2024) Seitkaliyeva G.
This research presents the development and implementation of a real-time focus tracking system in video streams, utilizing advanced computational methods and image analysis techniques. The core of the system is based on the L2CS-Net model, a convolutional neural network known for its efficacy in gaze estimation. The objective of this research was to use it for real-world applications, particularly in settings where user engagement and attention monitoring are crucial, such as e-learning and virtual meetings. To address the challenges of variable video quality and user behavior typically encountered outside laboratory conditions, the L2CS-Net model was refined using a comprehensive dataset, the MPIIGaze, with specific focus on preprocessing techniques that enhance training efficiency and model accuracy. The practical deployment of the model was achieved through a web-based interface, developed to provide real-time feedback on user focus. This interface was hosted locally on a Flask API, ensuring ease of deployment and making changes. A key feature of the system is an alert mechanism that notifies users when the detected gaze deviates beyond set thresholds, indicating potential lapses in attention. Experimental results demonstrate that the enhanced L2CS-Net model achieves high accuracy in gaze prediction. Furthermore, the web application’s real-time processing capabilities and the effectiveness of the alert system were validated under operational conditions. Future works can explore the integration of additional sensor data and improvements in alert accuracy, aiming to further enhance user engagement and monitoring in virtual environments.
ItemOpen Access
Traffic Sign Detection through Image Processing and Pattern Recognition
(Faculty of Engineering and Natural Science, 2013) Zhanibekov D.
Urbanisation, growth of cities and their population bring serious changes into our lives. That includes increasing numbers of cars on the road and traffic complexity. Since the very early stages of car and traffic development the very first concern of the designers and engineers was safety on the road. Most commonly it depends on the drivers and pedestrians directly. Attentiveness or inattentiveness of the traffic participants is one of the reasons why accidents can happen. Traffic signs provide important information for drivers about road condition and hazards. Their discriminating shape and colors make them easily recognizable by humans. Same factors can help development of a visionbased TSR system. Beside the application of TSR in autonomous vehicles, it can also serve as an assistant driver (e.g. when combined with speedometer output) to notify the driver about approaching a traffic sign (e.g. even before driver sees it) or his risky behavior (like driving above the speed limit). Safe driving was one of the three identified main work areas and meant to employ autonomous vehicle control for safer driving with less mental load on the driver. My proposed method is composed of three main Stages: 1. detection, which is performed using a novel application of maximally stable extremal regions 2. recognition, which is performed with LBP features 3. mobility, which is performed using mobile phone
ItemOpen Access
A thorough survey into the recognition of face emotion expression:experimental study, practical uses, and recommendations for the future
(Faculty of Engineering and Natural Science, 2024) Kuanyshbayev D.
The growth of the volume of information, as well as the expansion of the range of technically complex decision-making tasks require the systematization of existing methods and the development of new techniques and algorithms for their solution. The master’s thesis examines the possibility of using a neural network to solve the problem of recognizing human emotions. Artificial neural networks offer promising prospects for development, and software has a great advantage in using them. Moreover, each task performed has an unlimited and non-standard set of solution methods. The article considers the possibility of using a neural network to solve the problem of recognizing human emotions. The increasing volume of data, along with the breadth of technologically sophisticated issues with solving, necessitates the systematization of existing approaches and the creation of new techniques and algorithms for their resolution. The master’s thesis investigates the feasibility of utilizing a neural network to tackle the challenge of identifying human emotions. Artificial neural networks provide tremendous growth opportunities, and software can benefit greatly from their use. Furthermore, each challenge contains an infinite and non-standardized collection of solution techniques. The article discusses the feasibility of utilizing a neural network to tackle the difficulty of identifying human emotions.
ItemOpen Access
Analysis of students’ behavior and progress on Learning Management System using Machine Learning
(Faculty of Engineering and Natural Science, 2024) Kalekes D.
Many students do not put in sufficient effort at the beginning of the academic year, leading to grades that are insufficient for completing courses or obtaining scholarships. This study aims to analyze and predict student performance on the Moodle platform to provide early interventions and improve academic outcomes. The analysis focused on various courses from the 2023-2024 academic year at SDU University, selected due to their high average number of students and well-established structures. The research involved collecting data on three predictive factors: the number of completed assignments, the total time spent on the course, and the number of actions on the platform. Six machine learning algorithms were applied to predict student performance: k-Nearest Neighbor, Random Forest, Decision Tree, Logistic Regression, Naive Bayes, and Support Vector Machine. The study compared the effectiveness of early prediction at 5, 10, and 15 weeks into the courses. Key findings indicate that student activities on Moodle are significantly correlated with higher academic performance. The Support Vector Machine model showed the best results in the early weeks, while the Random Forest model demonstrated stable results over longer periods. These findings highlight the potential of machine learning models to identify at-risk students early, allowing for timely support and interventions. The implications of this research are significant for educators and administrators. The ability to predict student performance early can facilitate timely interventions, helping students improve their academic results and reduce withdrawal rates. This study contributes to the growing body of knowledge in educational data analysis and learning analytics, providing a foundation for future research to refine and expand predictive capabilities in educational institutions.