Browse
Recent Submissions
Item Open Access AI-Powered Approach to Career Path Prediction for IT Students Using Academic and Behavior Data(SDU University, 2025) Berilkozha B.A.The integration of educational outcomes with workforce requirements is now a pressing issue facing economies worldwide in the age of digital transformation.In addition to changing industries, the fast development of information and communication technology (ICTs) has also changed the skills that future professionals must possess [1]. Higher education institutions are under more and more pressure to equip graduates with the technical and adaptable skills necessary to meet the quickly evolving demands of the industry as global economies move toward automation, artificial intelligence (AI), and datacentric operations [2]. Of them, the field of information technology (IT) has become one of the most dynamic, requiring students to constantly adjust to challenging interdisciplinary problems, data-driven workflows, and new programming paradigms [3]. The choice of a career route is still quite difficult for university undergraduates, even with the growth of IT education. Finding job paths that fit their interests, abilities, and long-term goals is a challenge for many recent graduates. According to research, a sizable percentage of IT graduates wind up working in fields unrelated to their degree of concentration [4]. The absence of individualized career counseling programs that may combine behavioral and academic markers into a logical framework for decision-making frequently causes this mismatch. In the age of big data and artificial intelligence, traditional academic advising-which mostly depends on human intuition and small datasets-cannot scale efficiently [5]. Recent developments in machine learning (ML) and artificial intelligence (AI) have created new avenues for enhancing educational decision-making. Large amounts of behavioral and academic data can be analyzed by AI-driven systems to produce tailored recommendations for pupils [6]. This is in line with the larger worldwide movement toward data-driven education, which is referred to as Learning Analytics (LA) and Educational Data Mining (EDM) [7]. Whereas LA analyzes educational data to aid in institutional decision-making, EDM concentrates on identifying patterns in the data to improve learning outcomes. Predictive modeling has become well-known in this context due to its capacity to predict career outcomes, dropout risks, and student performance [8]. In the context of career counseling, artificial intelligence (AI) makes it possible to build prediction systems that determine the best career pathways for students based on their academic records, extracurricular activities, motivating factors, and personality features [9]. To interpret both structured and unstructured data, these systems use machine learning methods including Random Forest (RF), Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Gradient Boosting (GB), and Deep Neural Networks (DNNs). Research has indicated that models that employ ensemble techniques, such Gradient Boosting, frequently perform better than conventional classifiers because of their ability to manage intricate, nonlinear interactions between features [10]. However, there are also regional differences in how AI is incorporated into job counseling. Universities in developing nations, especially those in Central Asia, 7 are still in the early phases of adopting AI-powered academic support systems, whereas those in Europe, North America, and parts of East Asia have already done so [11]. Under the government's Digital Kazakhstan initiative, the creation of digital ecosystems in education has been designated as a strategic priority in Kazakhstan [12]. However, data-driven career advice is still completely unexplored despite significant improvements in digital infrastructure and e-learning platforms [Item Open Access Analysis and development of algorithm and method for object recognition(SDU University, 2025) Aitimov A.K.In an era characterized by the proliferation of visual data, the ability to comprehend and interpret the content of images and videos is of paramount significance [1]. Object recognition, a cornerstone of computer vision, plays a pivotal role in this endeavor. It constitutes the bedrock upon which numerous applications, ranging from autonomous vehicles to augmented reality systems, rely for accurate perception and decision-making. The fundamental task of object recognition is to endow machines with the capability to identify and categorize objects within a given visual context, akin to the cognitive abilities of the human visual system [2]. Pattern identification, description, categorization, and grouping provide challenges for Artificial Intelligence (AI), medicine, biology, and other branches of engineering and science. There are numerous definitions of the term "pattern". A pattern is a group of things, happenings, or thoughts that have certain similarities or features. According to Norbert Wiener, the essence of a pattern is an arrangement [3]. It is distinguished by the arrangement of its constituent pieces, rather than by their intrinsic characteristics the pattern is sometimes regarded as the opposite of chaos and a substantially namable item [4]. It can also be defined by a factor shared by multiple instances of the same object. Similarity in fingerprint pictures creates fingerprint patterns; handwriting, audio signals, web pages, and the human face are further examples of patterns [5]. Object recognition remains a challenging problem in computer vision due to various factors that affect the reliability and generalization of recognition systems. According to Bansal et al. in 2021, key challenges include the variability in object appearance, where differences in size, shape, color, illumination, and orientation make it difficult for models to learn consistent representations. Another major issue is scale and resolution, as detecting small objects within large and complex scenes requires algorithms capable of handling multi-scale information effectively. Partial occlusion also hinders detection accuracy when objects are partially hidden by others or by their own parts, while intra-class variability causes significant differences in appearance among objects of the same category. Moreover, Bansal et al. highlight that deep learning approaches often depend on large annotated datasets, and limited training data remain a major obstacle to achieving stable recognition performance [6]. Therefore, the present work focuses on addressing the limitations associated with insufficient data in object recognition tasksItem Open Access Profession inclination identification using machine learning(Suleyman Demirel University, 2021) Talasbek A.L.General characteristics of research. The given work is devoted to the research and development of an application that suggests recommendations for future profession selection based on the personal characteristics of a person by identifying professional inclinations. Relevance. Currently, the Kazakhstan market has virtually no systems for profession inclination identification. Modern society makes new demands on performance and professionalism. However, high levels of professionalism suggest a full disclosure of the potential of the individual, which is impossible without taking into account the personal characteristics of an individual. Many of the questionnaires conducted by organizations do not sufficiently define and describe the type of person for appointment, selection of personnel for certain special programs, and do not give a reliable result about the person in question whether the person will cope with certain official duties.Item Open Access Open Vocabulary Model for Kazakh Language using Deep Neural Networks(Suleyman Demirel University, 2021) Sultanova N.Zh.Assessment of the current state of the scientific and technological problem being solved. For the past 25 years there has been a demand for software solutions related to text processing, which has repeatedly experienced periods of growth, related to the emergence of personal computers, and with the rapid development of the Internet, and the rapid development of the Internet, and, In this natural language remains the most important way of communication, be the input of the search query on the miniature screen of the mobile phone, hints of the car navigator or business correspondence. Practically in all such applications such or otherwise the language model is used. So, for a convenient input of texts on a mobile phone, it is necessary to use the predicate input system, which practically corresponds to the direct application of the language model; language model - an indefinite part of the system of speech recognition, including volume and vocal search; Linguistic models are used in machine translation systems, the quality of which at the moment is still far from ideal, but still grows steadilyItem Open Access Development of methods, algorithms of machine learning for Kazakh speech recognition(Suleyman Demirel University, 2021) Kuanyshbay D.N.Automatic speech recognition system is on its way to reach its full potential mainly because of the significance of the field and rapid growth of the speech data. However, the growth of speech recognition for low-resourced languages like Kazakh is not observed, since there is not enough data and there is no automated tool for data collection. All existing speech recognition models for Kazakh language are based on datasets that were collected locally and manually, which makes these datasets private and inaccessible for other researchers to use them. Manual speech data collection requires a lot of time and effort, which may not match the quality requirements. Low quality speech data may ultimately affect the performance of speech recognition model considerably. This research work focuses on two important steps of building reliable automatic speech recognition system for Kazakh language – 1) construction of automated speech data collection system, 2) application of transfer learning to recurrent neural network. Automatic data collection system is based on a website, which perfectly segments the audio data and separates these audio files with corresponding transcriptions by speakers. As a result system produced around 100 hours of speech data, which in terms of structure is suitable for neural network to train. Transfer learning technique is based on Russian speech recognition model, which transfers all weights to the neural network that built for Kazakh language. Using transfer learning multilingual Automatic speech recognition model was obtained, which outperforms the simple LSTM based model by 32% and BiLSTM model by 24% (in terms of Label Error Rate).Item Open Access The Methodology of Using Active Learning Methods in Teaching Mathematical Analysis Courses to Students(2022) Abdullah AlmasRelevance of the work: The demand for specialists in mathematics and natural sciences should be said that is growing all over the world. Many governments and private organizations have modernized STEM education to effectively meet this demand and promote teaching to improve students’ mathematical skills. Recently, a lot of academics and organizations have emphasized that for students to succeed beyond graduation, they must acquire 21st-century skills. According to the study, the most important STEM talents are those that require collaboration, problem-solving, imagination, entrepreneurship, adaptability, critical thinking, initiative, effective community, access to information, analysis, and curiosity.Item Open Access Development of supervisorship system with tracking progress and the use of artificial intelligence(SDU University, 2024) Serek AzamatThe Supervisorship service has been developed with the feature to match students to supervisors based on psychological perceptions through multidimensional analysis of matching algorithms and the feature to track students’ progress. This study explores the utilization of four distinct algorithms for the purpose of student-supervisor matching. A comprehensive evaluation of these algorithms is conducted, encompassing a variety of metrics including preference satisfaction, workload balance, time and space complexities, minimum and maximum workload, and compatibility scores which this work introduced.Item Open Access Advisory system for adapting a single machine problem to a distributed solution(SDU University, 2024) Orynbekova KamilaGeneral characteristics of the work. The work encompasses developing an advisory system to recommend solutions for single-machine problems adaptable to distributed systems, mainly focusing on implementation within the MapReduce platform. Methodologically, an experiment evaluated learning effectiveness, while extensive data collection informed model development. Predictive models, including Naive Bayes and Logistic Regression, were optimized and integrated into a recommendation system validated through rigorous evaluation. The aim of the research is to develop an advisory system that recommends single-machine problem solutions that adapt to distributed systems and are suitable for implementation on the MapReduce platform.Item Restricted Research and Development of Intelligent Testing System based on Android Platform(Suleyman Demirel University, 2014) Bogdanchikov A.V.This thesis is devoted to developing advanced methods for intelligent testing system to test students’ knowledge and inspire prudent motivation to study. Set of methods are compared and stressed advantages and drawbacks of each, given different approaches in teaching students’ modern technologies. For some methods provided practical implementation on Android platform.