SDU Institutional Repository (SDU University)
Not a member yet
2006 research outputs found
Sort by
Fractal Dimension of Exceptional Sets in Semi-Regular Continued Fractions
No full textThis paper examines how the average value of the sequence bn in the Lehner expansion of a real number x influences its box dimension. Our primary objective is to analyze how variations in the average of bn impact the box dimension, which serves as a measure of the complexity of the sequence. Using the box-counting method, we numerically estimate the box dimension and explore its relationship with the fractal nature of Lehner expansions
Ұйым персоналының қызмет нәтижесін талдау және бағалау
Get PDFЗерттеудің өзектілігі – дәстүрлі, сирек және көбіне формальды бағалау рәсімдерінің орнына қызметкерлердің тиімділігін құндылыққа негізделген және үздіксіз кері байланысқа сүйенген заманауи тәсілдермен бағалау қажеттілігінен туындайды. Адами капитал бәсекелестіктің негізгі факторы болғандықтан, бағалау жүйесін жаңғырту ұйымның ұзақмерзімді нәтижелілігі мен кадр тұрақтылығын қамтамасыз етеді. Зерттеудің мақсаты – қызмет нәтижесін бағалау жүйесін құндылыққа негізделген заманауи әдістер арқылы жетілдіру жолдарын айқындау. Ғылыми жаңалық – құндылықтарға негізделген бағалау қағидаттарын Қазақстан контекстіне бейімдеп, дәстүрлі көрсеткіштермен интеграцияланған жүйелік модель ұсыну; білім беру саласындағы эмпирикалық деректермен негіздеу. Теориялық және тәжірибелік маңыздылық – модель түрлі ұйымдарда бейімдеуге жарамды «жол картасын» береді; нәтижелер кадр тұрақтылығын, engagement деңгейін және ұйым мәдениетін нығайтуға бағытталған басқарушылық шешімдерге негіз болады. Шектеулер – уақыт пен іріктеме көлемі, субъективті жауаптар ықпалы, бюджеттік шектеулер
Transforming the Research Commercialisation Ecosystem from Science Push to Open Innovation Model
No full textThis paper provides a critical analysis of the commercialization of research in a developing country that is undertaking science and technology policy reforms with an emphasis on the shift from a science-push to an open innovation model. Research commercialization is pivotal for technological advancement, but many innovations fail to reach the market, falling into the so-called "Valley of Death." In Kazakhstan, recent reforms are aimed at overcoming this gap, but the persistence of traditional linear models presents considerable obstacles. Through document analysis and semi-structured interviews, this study explores the nature of these reforms and the barriers encountered by academia and industry in the commercialization process. Theoretically, the study makes a contribution by examining the challenges of shifting from linear innovation models to more complex systems approaches. The findings underscore the difficulty of this transformation in Kazakhstan, highlighting the legacy of the Soviet-era focus on science-push models and the insufficient capacity of local industry to absorb new knowledge. The paper argues that building an effective commercialization pathway requires substantial investment in research infrastructure, capacity building, and regulatory adjustments to facilitate stakeholder collaboration
Comprehensive Analysis of a Recommender System for Career Guidance
Get PDFChoosing a suitable specialty is crucial for students, influenced by personal interests, academic performance, and career prospects. However, many struggle due to a lack of clear guidance and information overload. This research proposes a recommendation system to help students choose appropriate specialties and elective courses based on their academic performance and grades. The study focuses on IT students, as this field offers a wide range of specialties and career opportunities. It utilizes machine learning techniques, including reinforcement learning algorithms, to analyze academic data and provide personalized recommendations. The study compares traditional machine learning algorithms like Decision Tree, Support Vector Machine, and Random Forest with reinforcement learning algorithms such as Q-learning and Deep Q-Network. The methodology involves data collection, preparation, and feature engineering, followed by implementing various classifiers to build the recommendation system. Results indicate that Q-learning achieves the highest accuracy in recommending specialties, outperforming other algorithms. However, traditional machine learning algorithms also show competitive performance, suggesting both approaches can be effective. This research contributes to educational technology by offering a practical solution to help students make informed academic and career decisions. Future work includes enhancing the recommendation system with real-time data and user feedback mechanisms to improve its effectiveness and usability
Mathematical modeling of infectious diseases and the impact of vaccination strategies
Get PDFMathematical modeling plays a crucial role in understanding and combating infectious diseases, offering predictive insights into disease spread and the impact of vaccination strategies. This paper explored the significance of mathematical modeling in epidemic control efforts, focusing on the interplay between vaccination strategies, disease transmission rates, and population immunity. To facilitate meaningful comparisons of vaccination strategies, we maintained a consistent framework by fixing the vaccination capacity to vary from 10 to 100% of the total population. As an example, at a 50% vaccination capacity, the pulse strategy averted approximately 45.61% of deaths, while continuous and hybrid strategies averted around 45.18 and 45.69%, respectively. Sensitivity analysis further indicated that continuous vaccination has a more direct impact on reducing the basic reproduction number R0 compared to pulse vaccination. By analyzing key parameters such as R0 , pulse vaccination coefficients, and continuous vaccination parameters, the study underscores the value of mathematical modeling in shaping public health policies and guiding decision-making during disease outbreak
Gender inequality in the top management of Kazakhstan companies
Get PDFThis thesis investigates the problem of gender inequality in the top management of companies in Kazakhstan, with a focus on large and small companies. The main objective of the work is to identify the extent of gender inequality, analyse the factors contributing to this inequality and assess its impact on companies' productivity and employee morale. The study is based on qualitative data analysis. In a large company in Kazakhstan, a national oil and gas company, women hold only 10 per cent of top management positions. As part of this analysis, 50 employees were interviewed, of whom 20 per cent reported barriers to the promotion of women to senior positions. In comparison, in smaller companies in Kazakhstan, the average proportion of women in senior management is 25 per cent. Analysis of data from 30 small companies showed that 30% of employees believe that gender inequality is less pronounced in their company compared to large companies. The main factors contributing to gender inequality include socio-cultural attitudes, economic barriers and organisational characteristics such as lack of flexible schedules and lack of support programmes for women. The study found that gender inequality has a negative impact on company productivity and employee morale. In a large company where gender inequality is more pronounced, employees show 15% lower levels of job satisfaction compared to small companies. In conclusion, the thesis offers recommendations to reduce gender inequality, including introducing flexible working hours, developing support and mentoring programmes for women, and changing company culture and policies. In addition, recommendations are made for government agencies to create favourable conditions for achieving gender equality in senior management of companies in Kazakhstan
Automatization of object detection with AI
Get PDFThis study performs a comprehensive analysis of the YOLO (You Only Look Once) object detection method, painstakingly evaluating its performance on a wide range of image formats. The investigation’s main focus is on critical metrics that are carefully examined across set of different images, including processing time, frames per second (FPS), and important metrics related to object detection. By means of this rigorous examination, the research reveals noteworthy variations in the algorithm’s effectiveness, illuminating its intrinsic merits and demerits under various circumstances and situations.These results provide a vital source of information for practitioners and researchers working in the field of real-time object recognition applications. They enable them to make informed decisions and create optimization plans specifically for YOLO-based systems. This study provides stakeholders with the tools and considerations needed to efficiently negotiate the complexity of real-world deployments by providing a detailed understanding of the algorithm’s performance peculiarities, thereby promoting improvements and innovation in the field of computer vision
Numerical methods for matrix completion problem
Get PDFThis work addresses the significant challenge of framework completion in machi– ne learning, focusing on enhancing the accuracy and computational productivity of algorithms under conditions of large, noisy, and incomplete datasets. Central to this work are changes to two primary matrix completion techniques: Singular Value Thresholding (SVT) and Collaborative Filtering (CF). These methods were systematically progressed to handle common real-world data issues such as noise and sparsity, and were thoroughly tried over different applications, demonstrating significant execution enhancements. Through a detailed theoretical analysis, this research contributes robust frameworks for the convergence behaviors of these algorithms, giving a solid foundation for their application in practical scenarios. Improved SVT calculation, in particular, shows considerable reductions in Mean Absolute Error (MAE) and Mean Squared Error (MSE), indicating a superior performance over conventional methods. Besides, the refined CF approach presently integrates novel matrix factorization procedures, improving its utility in dynamic, personalized recommendation systems.The thesis underscores the potential of these refined algorithms in diverse fields, from advanced media to educational analytics, and sets a course for future investigate that incorporates integrating deep learning models and expanding into new data structures like tensors
Book text recognition in Kazakh Language
Get PDFThe digitization of Kazakh textual content poses unique challenges, particularly due to the language’s typographical diversity and the scarcity of digital resources. This thesis presents a novel approach to Optical Character Recognition (OCR) tailored to Kazakh book texts, leveraging a synthetic dataset to overcome the limitations of data scarcity and enhance model accuracy. Through meticulous dataset engineering, employing tools like SynthTiger, the study generates images that closely replicate the conditions of Kazakh printed material. The OCR models are rigorously trained and tested, demonstrating high precision in recognizing diverse text presentations. Additionally, this work includes the development of a web application utilizing the EasyOCR framework, which underscores the practical application of the research. Hosted on Hugging Face Spaces, the application offers users the capability to extract text from various image and document formats, illustrating the robustness and adaptability of the OCR models to real-world scenarios
Development of transcript-Driven IT Specialization Recommendation System using ML
No full textRecommendation systems in education are pivotal for guiding students through their academic and career paths. However, traditional systems often fail to address the unique challenges and rapid changes within the Information Technology (IT) sector. This study proposes a machine learning-driven approach to enhance the precision and personalization of IT career guidance.This research develops a sophisticated machine learning model using a variety of algorithms, including Random Forest, Logistic Regression and Decision Trees, to analyze and process detailed student transcripts. The study aims to predict and align students’ IT specializations with both their capabilities and market demands. A robust validation framework, including cross-validation and algorithm comparison, ensures the accuracy and reliability of the recommendation system. The model demonstrates a high degree of predictive accuracy, outperforming traditional recommendation systems. It effectively identifies individual strengths and market opportunities, providing tailored recommendations that improve educational outcomes and job market readiness. Integrating machine learning with educational recommendation systems offers a promising avenue for addressing the specialization needs within the IT sector. By leveraging detailed transcript data and advanced predictive analytics, the proposed system aligns educational paths with professional demands, enhancing student employability and meeting industry needs