Scientific Journal of Astana IT University
Not a member yet
250 research outputs found
Sort by
NUMERICAL SIMULATION OF WATER FLOW THROUGH A POROUS MEDIUM: VERIFICATION BY THE LIN 1999 EXPERIMENT
Get PDFThe presented study verifies a numerical model of fluid flow through a porous structure based on the experiment by Lin (1999). Water flows through porous media with a free surface are common in hydraulic engineering applications, such as dam breaks, seepage through dams, and the operation of wave protection structures. For a more accurate forecast, numerical modeling and verification should be performed using reliable experimental data. The experiment studied flow motion after a sudden removal of a partition (analogous to a dam break) in a rectangular channel with a porous obstacle. The laboratory setup had dimensions of 0.892 m × 0.37 m × 0.44 m, and the porous insert of 0.29 m × 0.37 m × 0.44 m was placed in a section of 0.3–0.59 m along the X-axis. Thus, the porous barrier blocked the cross-section of the channel, and water could flow only through its pores. This work helps to convey the forecast and allows to adequately simulate natural "jams" of branches and stones. This work demonstrates how, using such a verified model, it is possible to predict the flow dynamics in real conditions: water level changes, velocity field and coastal sediment accumulation zones. In addition, obtained data can serve as a basis for early warning of environmental risks and development of measures to protect water resources. In the future, it is planned to apply the model to a real section of the Talas River for a more detailed and reliable assessment of water pollution processes
LEVERAGING BIG DATA FOR DOG HEALTH ANALYSIS: AN EXPLORATORY STUDY USING "TANBA" IN KAZAKHSTAN
Get PDFIn the era of artificial intelligence, collecting and analyzing data about dog health through electronic medical cards and passports has become a key factor in improving the quality of life for pets. In this study there was analyzed 93,922 records about dogs contained in Kazakhstan's pet registration information system “Tanba”. The research focused on the demographic characteristics of dogs, including breed, age, and region of residence. Explanatory Data Analysis was conducted using descriptive statistics, and Natural Language Processing (NLP) methods were applied to standardize breed names, improving data consistency. Additionally, an ANOVA test was performed to assess the impact of factors such as gender, region, breed, and breed size on dogs' lifespan. Based on the data analysis, there are highlights of key aspects such as the predominance of young dogs (average age 5.52 years), the high proportion of dogs without breed, and the high concentration of stray animals in some regions, which emphasizes the need for increased efforts to control the population and improve living conditions for stray dogs. This study presents an analysis of the dog population for 2024 based on data from the Tanba national registration system. Unlike previous studies that focused on the prevalence of individual diseases or were limited to data from specific regions, this study covers the entire country and provides a general overview of the dog population. The findings indicate a high proportion of mixed-breed and stray dogs in Kazakhstan, as well as significant regional differences in canine lifespan. Breed and regional factors have a statistically significant impact on lifespan, emphasizing the importance of considering these characteristics when developing programs to improve animal welfare and veterinary care. In the future, it is planned to improve data processing algorithms and expand the use of additional sources of information, which will allow for more accurate assessment of dog health risks and development of more effective preventive measures
DEVELOPMENT OF A SOUND-BASED MOBILE APPLICATION FOR ROAD ACCIDENT DETECTION USING MACHINE LEARNING AND SPECTROGRAM ANALYSIS
Get PDFRoad accidents continue to pose a serious threat to public safety, underscoring the need for innovative, automated emergency response systems. This study presents the development of a mobile application that detects road accidents by analyzing audio signals in real time and immediately sends SMS alerts with GPS coordinates to emergency services and user-specified contacts. The system comprises two parts: a user-facing Android application and a server-side component for data processing. To build and train the detection models, we leverage the MIVIA Road Audio Events dataset and applied preprocessing techniques including amplitude normalization, background noise filtering, and data augmentation. Feature extraction involved zero-crossing rate, spectral centroid, spectral flux, energy entropy, short-time Fourier transform (STFT), and Mel-frequency cepstral coefficients (MFCCs). Two classification approaches were investigated: traditional machine learning models (Support Vector Machine, Random Forest, Gradient Boosting) and a deep learning model based on convolutional neural networks (CNNs) using Mel spectrogram inputs. Experimental results demonstrate that the CNN model achieved the highest performance with 91.2% accuracy, 89.5% recall, and an F1-score of 90.3%, outperforming the best classical model (Random Forest), which achieved 85.1% accuracy. The system also reduced the average accident alert time from 5–7 minutes to 1–2 minutes, representing a 60–80% improvement in emergency response speed. These results confirm the system’s reliability and practical benefit, particularly in regions like Kazakhstan, where timely medical intervention is critical. Limitations include reliance on smartphone availability, internet access, and environmental sound conditions. Future work will explore real-world testing, integration of accelerometer and gyroscope data, and deployment of edge computing for faster on-device processing. Overall, the proposed solution is a cost-effective, scalable approach for improving road safety and saving lives through rapid, automated accident detection
GLOVE-EMBEDDED ATTENTION BILSTM NETWORKS FOR ENHANCED MULTICLASSIFICATION OF TWEETS IN CYBERBULLYING DETECTION ON ONLINE CONTENT
Get PDFThis paper offers a neural network method for social media cyberbullying detection and classification. The model uses GloVe-embedded BiLSTM networks with self-attention to recognize language and semantic patterns. The research uses advanced machine learning methods to fight cyberbullying and suggests ways to improve cyberbullying detection systems' precision and ethics. The proposed paradigm addresses several cyberbullying levels and forms, enabling targeted interventions and victim support. GloVe implementations do semantic processing, BiLSTM networks sequentially learn, and self-attention mechanisms focus contextual analysis in the model. Word clouds show the abundance and relevance of phrases across several cyberbullying categories, revealing common themes and vocabulary. Tweet lengths, confusion matrix, training and validation loss and accuracy metrics, and ROC curves included in the dataset. The logistic regression model's ROC curve investigation shows substantial classification performance across multiple categories with AUC values between 0.905 and 0.997. The best model for age categorization has an AUC of 0.997, followed by religion (0.996) and ethnicity (0.993). Gender classification has an AUC of 0.979, whereas cyberbullying and non-cyberbullying have 0.921 and 0.905, respectively. The logistic regression model's ROC curve investigation shows substantial classification performance across multiple categories with AUC values between 0.905 and 0.997. The best model for age categorization has an AUC of 0.997, followed by religion (0.996) and ethnicity (0.993). Gender classification has an AUC of 0.979, whereas cyberbullying and non-cyberbullying have 0.921 and 0.905, respectively. The study encourages AI technology for social good and emphasizes the need to improve categorization algorithms to handle cyberbullying language's complex changes. Expanding training datasets, exploring hybrid modeling methodologies, and creating AI application ethics must be future goals
USING GRAPH CENTRALITY METRICS FOR DETECTION OF SUSPICIOUS TRANSACTIONS
Get PDFThis study addresses the critical challenge of detecting suspicious transactions in modern financial networks, focusing on the persistent threat of money laundering and related fraudulent activities. We propose a graph-based approach where each financial participant—whether an individual or an institution—is modeled as a node, and directed edges represent the flow of transactions. Using a dataset of anonymized banking records, we construct a directed graph and then calculate centrality measures, including degree centrality, betweenness centrality, closeness centrality, and eigenvector centrality. These metrics quantify how actively each node participates in or controls the circulation of funds across the network. Nodes characterized by particularly high values for betweenness or degree centrality emerge as potential “bridge” entities, acting as conduits for the majority of transaction paths. Our results indicate that these high-centrality participants may be key to understanding illicit financial flows, because they facilitate significant volumes of transactions or exert disproportionate influence by connecting otherwise separate sub-networks. Furthermore, a visualization of subgraphs around these nodes reveals tightly knit structures, suggesting the presence of possible hidden clusters that could be orchestrating complex money-laundering schemes. Overall, the proposed network-driven approach provides an efficient lens for early detection of suspicious accounts and transaction routes, especially when integrated with contemporary machine learning technologies for real-time analytics. The study concludes that centrality-based screening can enhance both the speed and accuracy of anti-fraud interventions, thereby strengthening the resilience of financial institutions in an increasingly data-rich and interconnected global economy
ADVANCED IMAGE COMPRESSION METHODS: A COMPARATIVE ANALYSIS OF MODERN ALGORITHMS AND THEIR APPLICATIONS
Get PDFThe paper examines in detail the modern methods of image compression, focusing on how advanced algorithms are used in practical digital imaging systems. The study examines many compression methods, including LZMA, LERC, ZSTD and their mixed forms and compares how well they perform in terms of compression ratio, time required, memory efficiency and how much information entropy they keep. Machine learning methods for compression are used in the analysis, focusing on how they work with images from medical imaging as well as satellite data. Experiments are performed on standardized datasets, with the main goal of following the theoretical limits set by Shannon’s Source Coding Theorem. The study shows that using modern hybrid algorithms, it is possible to compress data by at least 4:1 and keep it safe, with LZMA and LERC combinations performing best when the data is subject to entropic constraints. The results show that using parallel processing leads to a 60% decrease in processing time when compared to traditional single-threaded methods. The results strengthen the theories and techniques needed for the next generation of compression systems, mainly for handling high-resolution images quickly
RECOGNITION OF THE WATER SURFACE ACCORDING TO ICEYE DATA USING MACHINE LEARNING
Get PDFThe growing frequency of floods and the resulting socio-economic losses highlight the need for accurate and automated tools for detecting and monitoring water surfaces. This study presents a methodology for automatic water surface recognition based on high-resolution ICEYE synthetic aperture radar (SAR) data. The algorithm is implemented in the Google Earth Engine environment and uses the Random Forest machine-learning model trained on manually labeled “water” and “land” classes derived directly from the radar imagery. Preprocessing, performed in ESA SNAP, included radiometric calibration, Range-Doppler terrain correction, and speckle filtering to ensure accurate backscatter representation. The trained model was applied to ICEYE VV-polarized images acquired over Uralsk, Kazakhstan, on April 20–21, 2024, during a major regional flood. To validate the results, the Random Forest–derived masks were compared with those obtained using traditional methods such as Otsu and fixed-threshold classification, as well as optical masks generated from Sentinel-2 NDWI and MNDWI indices. Quantitative evaluation showed an overall accuracy of 76.8 % and a kappa coefficient of 0.535, while the area under the ROC curve (AUC = 0.91) indicated strong discriminatory capability. The Random Forest model demonstrated greater spatial precision and reduced false-positive mapping compared to threshold-based methods, confirming its suitability for operational flood monitoring. The proposed approach highlights the potential of ICEYE data for near-real-time water surface mapping, especially under cloud-covered conditions where optical sensors are ineffective. Moreover, the developed workflow ensures reproducibility and can be integrated into automated flood-response systems for rapid situation assessment. In the future, incorporating additional polarimetric and texture features is expected to further enhance model performance and extend its applicability to diverse hydrological environments
METHODS AND ALGORITHMS FOR SOLVING THE PROBLEM ON THE SUM OF SUBSETS
Get PDFWe study special-case algorithms for the subset-sum problem when the subset size is fixed to , using algebraic and geometric formulations that yield practical procedures with clear time and space bounds. The subset sum problem is one of the fundamental problems in computational complexity theory. It consists of determining whether, given a finite set of non-negative integers, there exists a subset whose sum of elements is equal to a predetermined number. This problem belongs to the class of nondeterministic polynomial time complete (NP-complete) problems: its solution can be verified in polynomial time, but an efficient algorithm for the general case has not yet been found. The goal of our research is to find new methods for solving the subset sum problem for special cases using algebraic and geometric approaches. The proposed method is based on a polynomial formulation of the problem inspired by Waring's conjecture for polynomials and the Neumann–Slater theorem. The main idea is to construct polynomials whose coefficients contain information about the sum of the elements of a subset. Using Vieta's theorem and the Euclidean algorithm, the problem is reduced to checking whether certain algebraic conditions are satisfied. The article proposes two lemmas proving the polynomial solvability of the subset sum problem for subset cardinality two and three. Based on them, two algorithms are developed: one uses value mapping and a fusion method, the other is based on a geometric criterion for collinearity of points obtained by transforming set elements. The algorithms demonstrate efficiency in terms of time and memory and do not require division into verification and decision stages. Effective methods for solving it allow us to develop faster algorithms for intelligent information processing, optimization of computing processes, and construction of reliable data protection systems. Our results establish polynomial-time solvability only for these fixed-???? cases and do not claim consequences for the general subset-sum problem or for the P vs NP question
DEVELOPMENT OF TIME SERIES FORECASTING MODELS FOR AIR POLLUTION BASED ON DEEP SPARSE TRANSFORMER NETWORKS
Get PDFThis study investigates the application of fractal analysis and deep learning methods for forecasting pollutant emissions from the Ekibastuz coal-fired power plant. The research is based on time series of NO, NO₂, and PM₁₀ concentrations collected by industrial sensors during 2023–2024. To assess long-term dependencies, an R/S analysis was performed, and the results demonstrated stable persistence with average Hurst exponent values exceeding 0.67. This confirmed the appropriateness of employing models capable of capturing long-range memory in the data. In the second stage, a Deep Sparse Transformer Network (DSTN) architecture was implemented and adapted to the task of emission forecasting under different boiler operating modes. DSTN combines the advantages of transformer-based models with a sparse attention mechanism, which reduces computational complexity and enables efficient handling of long sequences. The model was trained using the PyTorch framework on a dataset of more than 67,000 records, with forecasting performed at horizons of 1, 6, 12, and 24 steps. The highest accuracy was achieved for short-term forecasts: the coefficient of determination for NO₂ reached 0.95 at a one-step horizon and decreased to 0.38 at 24 steps. For NO and PM₁₀, R² values ranged from 0.93 to 0.26. These findings indicate that DSTN is a highly effective tool for short-term forecasting but less accurate at longer horizons due to error accumulation. The results confirm the practical value of integrating fractal analysis with transformer architectures for emission monitoring and coal power plant operation management. The proposed approach can be embedded into industrial control systems to enable timely responses to peak emissions, optimize combustion modes, and mitigate environmental risks
INTELLECTUAL HARDWARE-SOFTWARE COMPLEX FOR FIBER-OPTIC SYSTEM MONITORING WITH CLASSIFICATION OF THE EVENTS AND RECOMMENDATIONS
No full textCurrently, there are many different methods of monitoring extended facilities. However, the most accurate, efficient, and more accessible methods are using fiber-optic sensors. This study examines existing methods based on the application of optical time-domain reflectometry (OTDR). Data from three main databases, namely Web of Science, Scopus, and Google Scholar, were considered as existing solutions. Among the existing types, the possibility of using interferometers was also taken into account. However, such systems are expensive and very sensitive. At the same time, OTDR systems have huge disadvantages, such as the relatively low sensitivity of such systems, the closeness of the solution, and the lack of integration. However, all the disadvantages, except for the proprietarity, can be eliminated by using a neural network. Therefore, a system based on an open architecture is proposed with the possibility of application on new and already installed monitoring systems using a neural network for classification and an expert system for assessing the situation and recommendations for the implementation of restoration work. A universal intelligent hardware–software complex is proposed, which includes modules for signal preprocessing based on Fourier transform, statistical filtering using the three-sigma method, event classification, and interpretation. The suggested developed system enables noise suppression, event recognition (vibration, bending, cable breakage), and generation of recommendations through artificial intelligence. A convolutional neural network was used as a neural network for event classification. Recommendations and evaluation were provided using an expert evaluation module based on the use of Copilot, which reduces decision-making time and prevents possible breakdowns