TELECOMMUNICATION AND INFORMATION TECHNOLOGIES

Title

2025-06-26T06:34:30+00:00

Scientific Journal

No. 2 (87) 2025

Content

2025-06-26T06:34:36+00:00

Content

METHOD FOR PROTECTING INFORMATION FROM LEAKAGE THROUGH PHYSICAL CHANNELS WITH SPECIFIED SECURITY PARAMETERS

2025-06-26T06:34:31+00:00

The article studies the problem of information leakage through the material channel. Such threats can arise as a result of intentional actions of attackers or unintentional leaks associated with the human factor, technical features of information carriers and imperfection of organizational security measures. It is shown that not too many publications are devoted to the study of the material channel of information leakage in the world scientific literature. Existing methods for building effective protection systems should be adapted by determining the permissible parameters of the number of attempts by an attacker to penetrate the object of information activity and the permissible time that can be allocated for such penetration. The article substantiates the theoretical foundations of building a system for protecting information from leakage through the material channel. The author proposes a method for protecting information from leakage through the material channel with specified security parameters, which takes into account the number of permissible penetration attempts and the time that an attacker needs to spend to penetrate the object to seize a technology sample or document. In the developed model, increasing the security parameter of the object leads to a decrease in the probability of penetration. The method is based on determining the line of probability of penetration and allows not only to design effective protection systems, but also to control the penetration process itself by the coincidence or deviation of the security incidents that have occurred from the line of penetration. Modeling such an approach allows you to predict the direction in which the designed process of penetration into the object is going, and to analyze it. It is concluded that further research on the construction of effective information protection systems against leakage through a material channel can be focused on methods for determining the probabilities of penetration and indicators of the security of protection systems, taking into account the characteristics of specific organizations.

Keywords: cybersecurity, information protection, information leakage, technical information leakage channel, material information leakage channel, probability of penetration.

HYPERSPECTRAL ANALYSIS FOR RECOGNITION AND CLASSIFICATION OF EARTH'S SURFACE MATERIALS

2025-06-26T06:34:36+00:00

This paper analyzes modern approaches to hyperspectral image classification, including machine learning methods that demonstrate high efficiency. Hyperspectral analysis enables precise identification of Earth's surface materials by utilizing the spectral characteristics of objects. However, hyperspectral image classification remains challenging due to the high dimensionality of data, the lack of labeled samples, and significant computational costs.
The authors examine the XGBoost and Random Forest algorithms, which are used for accurate material recognition based on their spectral properties. These methods allow the development of an effective classification system of Earth’s surface objects, while reducing computational costs and maintaining recognition accuracy. A comparative performance analysis of the two selected models is conducted based on metrics such as Confusion Matrix, Accuracy, Sensitivity, Specificity, Precision, and Recall.
Particular attention is paid to the problem of dimensionality reduction. Such data compression is a crucial step in hyperspectral image classification. The application of appropriate methods helps reduce the computational complexity and improves algorithm performance. Additionally, the study explores the adaptability of classification models to different datasets, which is essential for their practical application in the real-world scenarios.
The research results show that the proposed classification system ensures accurate identification of Earth’s surface materials. The findings can be applied in the military intelligence for object detection, environmental change monitoring, and the recognition of potentially hazardous materials. The proposed methods and models enhance the efficiency of hyperspectral data analysis, providing new opportunities for further research in this field.

Keywords: Hyperspectral analysis, ROSIS, XGBoost, RandomForestClassifier, classification task, image recognition.

EFFECTIVE SOLUTIONS FOR RAPID DETECTION OF COMMITTED PCS IN THE INFOCOMMUNICATION NETWORKS

2025-06-26T06:34:41+00:00

Recently, the need to solve problems in conditions of limited time resources has become increasingly relevant. This could be, for example, a network attack on a company's corporate resources, as a result of which an unknown number of PCs have been compromised, while being completely ignored by AV and IPS, which in turn imposes a significant time constraint, since it is necessary to quickly establish the "degree of infection" of each individual PC and isolate it from other computers in the infocommunication network. In this case, the traditional forensic analysis of digital traces collected using Forensic Triage will be too long. When under normal conditions, these tasks were performed by well-known programs and this time was enough, now the question arises, how to speed up the "execution" of tasks if there are no competitive analogues for the program? Therefore, there is a need for additional capabilities that would reduce the program for extracting digital artifacts execution time while maintaining sufficient efficiency. Or reduce the time an IT analyst spends analyzing one PC, which will allow him to check more PCs per unit of time. The purpose of the study is to substantiate effective solutions to reduce the time an information security analyst spends on identifying a specific PC in the infocommunication network as infected/not infected. The work identifies a component/tactic without which modern computer viruses usually do not work. A list of programs for rapid virus detection and an optimization script using a relational table of artifacts are proposed, which allow reducing the number of elements required for further research by more than ten times. This helps IT analysts significantly save time on detecting the infection of a specific PC in the infocommunication network as "infected/not infected".

Keywords: information; cybersecurity, computer forensics, computer digital artifacts, Windows, MITRE, autostart, viruses.

ANALYSIS OF FEATURE IMPORTANCE AND EFFECTIVENESS OF MACHINE LEARNING MODELS IN PREDICTING TUBERCULOSIS CASES IN INDIA

2025-06-26T06:34:45+00:00

Tuberculosis (TB) remains one of the most severe infectious diseases globally, with India bearing the highest burden according to the World Health Organization (WHO). High population density, unequal access to healthcare, socioeconomic conditions, and comorbidities such as diabetes create a conducive environment for TB spread. This study analyzes the importance of factors influencing TB incidence in India and evaluates the effectiveness of machine learning models in predicting cases. The aim is to identify key determinants of TB spread and develop evidence-based recommendations to reduce the epidemiological burden in the region.
The study utilizes data from 2019–2022, sourced from open databases, including WHO and Indian government reports. The dataset comprises 126 records and 25 variables, encompassing diagnostic indicators (e.g., detected TB cases, multidrug-resistant TB, TB-HIV coinfection), social factors (e.g., tobacco and alcohol use), healthcare infrastructure, and treatment outcomes (e.g., success, mortality, treatment interruption). The analysis employed descriptive statistics, correlation analysis, multiple linear regression with L1/L2 regularization (Ridge, Lasso), and nonlinear machine learning methods, including Decision Tree, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Random Forest. Model accuracy was assessed via cross-validation using R² (coefficient of determination) and MSE (mean squared error) metrics.
Correlation analysis revealed no strong linear relationships between factors and total TB cases, suggesting nonlinear dependencies. Multiple linear regression showed low explanatory power (R² ≈ 0.3), while regularized methods (Lasso with α = 0.01) slightly improved generalization (R² = 0.1007). Among linear models, factors related to case notifications by gender and multidrug-resistant TB diagnosis were most significant. Nonlinear models proved more effective: initial analysis indicated Random Forest (R² = 0.4595 on test data) outperformed KNN and SVM, while Decision Tree suffered from overfitting (R² = -0.3044 on test data). To enhance accuracy, a new target variable—normalized TB cases per 100,000 population (total_inf)—was introduced, accounting for state population sizes. This adjustment significantly improved model performance: Decision Tree achieved R² = 0.8724, and Random Forest reached R² = 0.8378 on test data. Factor analysis confirmed that multidrug-resistant TB diagnosis (MDR/RR TB Diagnosed) and treatment center infrastructure (PMDT-Infrastructure) are key predictors, highlighting the critical role of medical resources and timely detection of resistant strains.

Keywords: tuberculosis, India, machine learning, multiple linear regression, random forest, decision tree, SHAP analysis, socioeconomic factors, multidrug-resistant tuberculosis, incidence prediction.

MATHEMATICAL MODEL OF DATA COLLECTION AND PROCESSING FOR A RECOMMENDATION SYSTEM FOR FORMING AN INDIVIDUALIZED EDUCATIONAL PROGRAM

2025-06-26T06:34:53+00:00

The effectiveness of recommendation systems, which are used in various fields and, in particular, in the educational system, is a relevant issue for research by scientists and software system developers. Using the example of a system that provides recommendations for selecting educational components of a curriculum based on various parameters, including previous academic achievements of students, data on their past course selections, and the choices of other students, etc. The main goal of the research is to develop a mathematical model capable of effectively predicting grades for courses that are offered to the student as optional.
In the current implementation, ML.NET tools are used, which automate the process of matrix factorization and loss function minimization, simplifying the development and tuning of the model. The main stages of implementation include training the model using the MatrixFactorizationTrainer, defining the trainer's parameters, and optimizing the model through gradient descent and regularization by Frobenius Norm.
There are also some directions for improvement, such as integrating content-based filtering, which will allow us to take additional course characteristics, such as type or complexity, into account in the future, significantly improving the accuracy of recommendations. Alternatively, the calculation of similarity between courses or students can be enhanced, which would allow for making recommendations to end users based on the similarity.
In conclusion, an approach is demonstrated that allows for creating an effective recommendation system that can be used to predict student performance and form personalized educational programs. Improving the accuracy of such systems opens up new opportunities for enhancing the educational process, fostering a more individualized approach to learning.

Keywords: recommendation system, matrix factorization, grade prediction, mean square error, root mean square error, Frobenius Norm, ML.NET, C#, content-based filtering, course selection, content-based filtering, education, student, individual study plan.

METHODS OF GDPR FOR ENSURING DATA STORAGE SECURITY AGAINST LEAKS AND THREATS

2025-06-26T06:34:57+00:00

This article explores the issue of protecting data storage systems from leaks and threats in the context of implementing the security requirements of the General Data Protection Regulation (GDPR). This topic is critically relevant given the increasing volumes of processed information. The authors emphasize the lack of specialized studies that examine the compliance of data storage security with GDPR standards. It has been determined that nearly all digital platforms, organizations, and institutions rely on data storage solutions, making them highly vulnerable to various threats—both external (hacker attacks, phishing, malware) and internal (human factor, intentional or unintentional leaks by employees of institutions or organizations).
This study systematizes five main categories of threats to data storage security: unauthorized access, internal data leaks, phishing attacks, malware, and issues related to improper data storage and backup management. For each category, real-world examples of security breaches from the past five years are provided, illustrating the impact of these threats on the protection of personal data. The article also examines the application of security measures such as data encryption (AES-256, TLS), access control, monitoring, pseudonymization, anonymization, backup strategies, and the automation of data processing, in accordance with various articles of the GDPR.
This research holds not only analytical but also practical value, as it provides a comprehensive perspective on how to implement both technical and organizational security measures for data storage in compliance with GDPR requirements. The study describes models for enforcing security policies, logging mechanisms, and data integrity verification. Additionally, it underscores the importance of fostering a culture of information security and raising awareness among employees.

Keywords: GDPR, data security, information leakage, data storage, encryption, threats, cybersecurity, attacks.

IMPLEMENTATION OF A SOLUTION BASED ON MICROSOFT AZURE INFRASTRUCTURE FOR AGRIBUSINESS MANAGEMENT

2025-06-26T06:35:02+00:00

The introduction of modern cloud technologies based on artificial intelligence (AI) in the agro-industrial complex (AIC) is not only an innovative trend but also a prerequisite for its development and overcoming the challenges of today, given the constant growth in demand for food and the need to preserve natural resources. AI technologies are becoming a critical tool for ensuring sustainable development and efficient management of the agricultural sector. They also contribute to the growth of productivity and accuracy of management decision-making and enable rapid response to the rapid changes in the farm sector due to the digital transformation of this industry and the economy as a whole. The article presents typical machine learning (ML) tasks that can be applied in the agricultural sector to improve business processes. Based on the Microsoft Azure cloud platform, an architecture has been built that provides modularity, scalability, flexibility, security, and integration of various components to optimize business processes in the agro-industrial complex and effective management in general. Azure Machine Learning is used to create and deploy forecasting and optimization models, which will help representatives of the agricultural sector to adapt to changes in climate conditions and weather events quickly, predict yields, field conditions, market prices, etc., and provide flexible management of business processes through interactive dashboards and APIs (RESTful API settings). In addition, recommendations for further improvement of the forecasting solution in the agricultural sector using cloud technologies by small farms with limited resources are provided. The implementation of the proposed solution based on Microsoft Azure infrastructure will be a valuable resource for professionals involved in the digital transformation of the country's agricultural sector, developers of cloud solutions to support various business processes in the agro-industrial complex, in particular for the implementation of typical machine learning (ML) tasks.

Keywords: cloud infrastructure, artificial intelligence, machine learning, agricultural industry, digital transformation, big data.

MODELS OF BLOCKCHAIN TECHNOLOGIES APPLICATION IN HIGH-LOAD COMPUTER SYSTEMS FOR DISTRIBUTED DATA STORAGE AND PROCESSING

2025-06-26T06:35:07+00:00

This study presents a comprehensive analysis of the application of blockchain technologies in high-load computing systems for distributed data storage and processing. Particular attention is paid to evaluating architectural approaches for integrating blockchain into modern distributed computing environments, including the use of smart contracts, consensus mechanisms, and hybrid storage models. The study highlights the key advantages of blockchain infrastructure, such as decentralization, data integrity assurance, enhanced security, and transaction transparency.
The research provides an in-depth review of contemporary blockchain-based distributed storage solutions, including IPFS (InterPlanetary File System), Filecoin, and Storj, along with their adaptation to corporate and cloud platforms. A special focus is placed on the challenges of blockchain scalability, transaction throughput, and efficient utilization of computational and energy resources, which are crucial for its implementation in high-load systems. Additionally, the study explores the impact of blockchain adoption on overall system performance and potential trade-offs between security, efficiency, and resource consumption.
Through an extensive review of existing research and practical implementations, the study identifies the core benefits of blockchain for distributed storage and processing. These include resilience to system failures, elimination of single points of failure, and protection against unauthorized data access. At the same time, several critical challenges are outlined, such as high energy consumption, the need for optimizing consensus algorithms, and developing effective strategies for managing large-scale data volumes. The research also investigates the economic implications of blockchain deployment, assessing cost-effectiveness in different operational contexts.
Based on the findings, the study proposes conceptual strategies for optimizing blockchain utilization in distributed computing systems. These strategies include the implementation of hybrid data processing models, integration with cloud computing, and enhancements in transaction management mechanisms. The insights provided in this study may be valuable for blockchain solution developers, researchers in distributed computing, and cybersecurity professionals, offering a robust foundation for further advancements in the field.

Keywords: blockchain, distributed storage, data processing, high-load systems, security, scalability, smart contracts, consensus, decentralization.

КОМПЛЕКСНИЙ ПІДХІД ДО ІНТЕЛЕКТУАЛЬНОГО УПРАВЛІННЯ, МОДЕЛЮВАННЯ ТА ВИЯВЛЕННЯ МЕРЕЖНИХ АНОМАЛІЙ НА ОСНОВІ ЕНТРОПІЙНИХ ТА НЕЙРОМЕРЕЖЕВИХ ПІДХОДІВ

2025-06-26T06:35:12+00:00

Abstract. This article presents a methodology for evaluating the reliability of information systems using systems analysis and entropy-based approaches. The proposed method is grounded in the formalization of diagnostic markers classified into three types: behavioral, structural, and performance-related. Within the study, an entropy-based reliability metric, RH, was developed to quantify the uncertainty level of a system’s state in realtime. Implementation variants using sliding windows and fixed intervals are proposed, providing high accuracy in failure prediction. Experimental modeling was conducted using data generated in the NetSim environment for a 200-node network. It was found that when RH drops below 0.6, the probability of failure within the next 30 minutes exceeds 85%. Additionally, the study examines correlations between different categories of markers and finds that combinations of structural and behavioral signals have the highest predictive potential. Heatmaps and histograms were generated to visualize the results, enabling integration of the model into decision support systems. Special attention was given to the construction of temporal activity profiles of network components and their impact on entropy indicators. The paper also evaluates the adaptability of the model to highly dynamic environments, such as cloud and distributed IoT architectures. Algorithmic approaches to automatic recalibration of system parameters under changing configurations or workloads are proposed. The research results demonstrate the high effectiveness of combining entropy analysis with marker-based approaches to enable proactive monitoring, early failure detection, and enhancement of cyber-resilience in complex information systems operating in unstable digital environments. The potential application of the proposed model is considered for industrial control systems, banking IT infrastructures, critical digital security facilities, and integration with neural network technologies.

Keywords: information systems reliability, entropy assessment, diagnostic markers, failure prediction, RH metric, heatmaps, sliding window, structural faults, NetSim, automatic response.

AIOT SYSTEM BASED ON GRAPH THEORY FOR THE MEDICAL FIELD

2025-06-26T06:35:16+00:00

Abstract: This study presents a comprehensive mathematical model for applying graph theory in AIoT (Artificial Intelligence of Things) systems designed for health monitoring. The model accounts for the structure of the sensor network, the temporal dynamics of physiological parameters, and the integration of intelligent data processing methods based on graph neural networks. IoT devices (such as heart rate monitors, glucose sensors,

channels. The developed model includes an adjacency matrix, vector representations of sensor parameters, and forecasting mechanisms using GCN and Temporal GNN architectures. Special attention is given to anomaly detection algorithms based on deviations between actual and predicted values of physiological signals. Mathematical approaches to threshold interpretation of these deviations are proposed, allowing timely identification of potential health threats. The study also considers dynamic optimization of graph structure through adaptive learning from input data, enabling the system to refine inter-sensor dependencies in real time. The findings of this research can be applied in the development of intelligent health monitoring systems for clinical settings, remote patient care, and mobile health solutions aimed at chronic disease prevention. The proposed model forms a foundation for the advancement of adaptive, personalized AIoT platforms in healthcare, capable of self-learning, forecasting, and real-time decision-making.

Keywords: Internet of Things, architecture, optimisation, automation, sensor, modelling, information system, network, graph theory, artificial intelligence.

MODEL OF AI INTERACTION WITH IOT DEVICES AT THE SOFTWARE UPDATE LEVEL

2025-06-26T06:35:20+00:00

Abstract: Modern IoT systems consist of millions of heterogeneous devices that require regular software updates to maintain functionality and security. Traditional methods of deploying updates are often insufficiently flexible and efficient, especially in dynamically changing network environments. This paper proposes an innovative model of interaction between artificial intelligence (AI) and IoT devices at the software update level that combines the advantages of canary releases with intelligent optimisation algorithms. The study focuses on the development of a mathematical model for intelligent update deployment that takes into account the following key aspects: phased implementation of changes (canary releases), automated monitoring of system stability, and adaptive decision-making based on machine learning algorithms. The proposed model uses Reinforcement Learning methods to dynamically select the optimal deployment strategy, which minimises the risk of failures and optimises the use of network resources.

Particular attention is paid to the formalisation of the deployment process using a mathematical apparatus, including: the definition of criteria for successful updates, cost functions for assessing the quality of deployment, and real-time decision-making algorithms. The results of the study demonstrate that the integration of AI into the process of updating IoT devices can significantly improve system stability and reduce the number of failures during large-scale deployments. The article also discusses promising areas for further research, including the introduction of federated learning for decentralised data analysis, integration of blockchain technologies to improve update security, and development of adaptive algorithms for heterogeneous IoT networks. The proposed approach opens up new opportunities for the creation of intelligent, self-managed IoT systems capable of operating effectively in a dynamically changing environment.

Keywords: artificial intelligence, IoT, information system, software updates, canary releases, machine learning, update deployment, network optimization.

MODELING OF CRITICAL STATES IN SIEM SYSTEM BASED ON CATASTROPHE THEORY

2025-06-26T06:35:25+00:00

Abstract: A study of the impact of cyber incidents on military information security management systems during the training of military units in training centers was conducted. An analysis of scientific research on the detection of cyber incidents using machine learning methods, which have their advantages and important disadvantages, has been carried out. It was found that the works do not consider the issues of system stability and the forecast of critical transitions of information system security states. The main advantages of using a SIEM system, which allows collecting, aggregating, storing and correlating events generated by a managed infrastructure, were determined. It was found that SIEM systems have significant disadvantages, including malfunction that prevent the detection of important threats, and no prediction of the development of events is carried out, which does not allow assessing future risks. An analysis of the main disadvantages of the SIEM system was conducted and solutions were proposed using a block with the catastrophe theory in the SIEM system. An analysis of modern systems for simulating the dynamics of combat operations in the format of command and staff training exercises in real time has been carried out. The structure of the integrated training system has been determined, as well as the main logical blocks that should be combined using SIEM into a single chain of events. The main components of the integrated training system, their purpose and the role of the SIEM system for responding to cyber incidents to establish data security have been established. An algorithm for detecting unstable system states during cyber incidents using SIEM and catastrophe theory in the integrated training system has been developed, which allows predicting and detecting unstable system states, as well as responding to information leaks in real time, which ensures an increase in the level of cyber resilience of the system.

Keywords: information security management system (ISMS), SIEM system, critical states, catastrophe theory, bifurcation points, cyber incident.

ALGORITHMICAL APPROACHES TO ANOMALIES DETECTION BASED ON MACHINE LEARNING

2025-06-26T06:35:29+00:00

Abstract. Anomaly detection in cybersecurity is a critically important process aimed at identifying atypical patterns of behavior or activity that significantly differ from established, normal operating procedures within an information system or computer network. These deviations from the norm can serve as early indicators of potential security threats, ranging from unauthorized intrusion attempts and malware distribution to exploitation of existing vulnerabilities in software or system configuration. Timely and effective detection of such anomalous events provides an opportunity to respond quickly to threats, prevent their further development, and minimize potential risks to the confidentiality, integrity, and availability of critical data and the organization’s digital infrastructure.

Modern cybersecurity anomaly detection systems increasingly use sophisticated machine learning algorithms to effectively recognize complex and subtle patterns of behavior. Unlike traditional methods that rely heavily on predefined rules and static thresholds, machine learning algorithms have the ability to learn from large amounts of data, which allows them to detect new and previously unknown types of attacks that static rules may not be able to handle. In cases where cybercriminals are constantly improving their methods and tools, the ability of machine learning algorithms to adapt to new threats by analyzing data on previous attacks and normal behavior becomes extremely valuable. These algorithms can detect subtle deviations that may be missed by systems based on strict rules, thereby increasing the overall effectiveness of intrusion detection and data leakage prevention systems. The use of machine learning allows you to build more intelligent and proactive security systems that can effectively counter modern cyber threats.

Keywords: anomaly detection, machine learning, information security, Z-score, statistical methods, neural networks, cyberattacks.

HYBRID APPROACH TO COMBINING NB-IOT WITH LORA

2025-06-26T06:35:34+00:00

Abstract: This work is dedicated to the comparative analysis of radio technologies for the Internet of Things (IoT) networks with the aim of identifying optimal solutions for ensuring reliable, energy-efficient, and scalable communication between devices. Since IoT technologies are used for data collection and exchange in dynamically changing networks, an important aspect is their ability to adapt to various environmental conditions and provide stable operation under resource constraints. The work examines the main wireless radio technologies used in IoT systems, including Wi-Fi, Bluetooth Low Energy (BLE), ZigBee, Z-Wave, LoRa, Sigfox, and NB-IoT. For each of these technologies, an analysis is performed of characteristics such as range and data transfer speed, energy efficiency, bandwidth, as well as the cost of deployment and infrastructure maintenance.

One of the key aspects is the comparison of these technologies in the context of their application in different IoT scenarios, such as urban and remote areas, as well as specific needs within critical infrastructures. Special attention is given to the potential integration of technologies like NB-IoT and LoRa within a hybrid approach, which combines the benefits of stable mobile connectivity and high energy efficiency. Such integration opens up opportunities for the development of multi-protocol solutions that ensure more efficient and reliable data transmission in low coverage-density conditions, particularly in rural and remote areas where traditional mobile networks cannot provide the necessary coverage.

The results of this research will contribute to the formulation of recommendations for the practical implementation of IoT systems, enhancing their performance and reliability, which can have significant potential for application in smart cities, industry, agriculture, and healthcare. The development of hybrid IoT networks, utilizing multiple different radio technologies, allows for a balance between cost-efficiency, device autonomy, and service quality, which could be key to the successful development of the Internet of Things in the future.

Keywords: Internet of Things, IoT, comparative analysis, radio technologies, hybrid architecture, LPWAN, WPAN, energy efficiency, NB-IoT, LoRa, IoT networks.

PROTECTED INTELLIGENT SYSTEM FOR AUTOMATIC NOTIFICATION OF IMPORTANT NEWS

2025-06-26T06:35:39+00:00

Abstract: The article provides a comprehensive analysis of existing systems of news portals, news aggregators and communication platforms, such as Telegram channels and Viber chats. It has been determined that, despite the access to a wide range of information sources, these solutions have a number of shortcomings: the need for independent monitoring of news, a high probability of information overload and limited filtering of really important events. To eliminate these shortcomings, it is proposed to create an automated system for notification of important news in real time without the participation of moderators. The use of machine learning algorithms, in particular artificial intelligence models such as ChatGPT, makes it possible to automatically analyze news texts, determine their importance and rank information according to specified criteria.

The article proposes the structure of a system for automatic notification of important news for its software implementation and substantiates the choice of a cross-platform application architecture based on Progressive Web App (PWA), which ensures the accessibility of the system to a wide range of users from different devices - smartphones, tablets and personal computers. The developed system consists of two main parts: a server one, which collects, analyzes and processes news data, and a client one - in the form of a convenient PWA application with support for offline mode, push notifications and the ability to install it on the device's home screen. To implement the server part, it is proposed to use modern programming languages, such as Node.js, Python or Golang, which guarantee flexibility, high performance and easy integration with external AI services via API. Data protection mechanisms, such as authorization and authentication via JWT, information encryption via HTTPS (SSL/TLS), database protection and secure caching in the browser, have been implemented. The proposed solutions guarantee confidentiality, integrity and availability of information. The testing has confirmed the high efficiency, accuracy of the system and convenience for end users.

Keywords: news texts, data collection, analysis, artificial intelligence models, web application, users, push notifications, AI services, security, database.

METHODS OF FORECASTING EXCHANGE RATES: ANALYSIS AND EVALUATION

2025-06-26T06:35:44+00:00

Annotation. The study examines key methods for forecasting exchange rates, including traditional approaches such as fundamental and technical analysis, as well as modern mathematical models, particularly ARIMA and GARCH. A comprehensive comparative analysis of the effectiveness of these methods under different economic conditions is conducted, taking into account the influence of macroeconomic indicators, political stability, global economic trends, and specific characteristics of local markets. It is shown that traditional approaches tend to lose their effectiveness under conditions of increased market volatility and crisis phenomena. This highlights the necessity of employing advanced data analysis methods, particularly machine learning algorithms, to enhance forecast accuracy and model adaptability.

Special attention is devoted to the development and implementation of a modified exchange rate forecasting method (MNKY), which combines classical econometric tools with artificial intelligence capabilities. The proposed method demonstrates high flexibility in model parameter adjustment, simplified integration of new data, and the ability to self-adapt to changing market conditions without compromising forecast quality. Experimental testing based on the UAH/USD exchange rate, using data from the National Bank of Ukraine, revealed that the average deviation of the MNKY method forecast was 2.5 units, significantly outperforming the ARIMA (8 units) and GARCH (4.8 units) models.

The study also reveals that integrating machine learning enables flexible consideration of the multifactor influence of political, economic, and social factors, traditionally complicating the forecasting of exchange rate dynamics. The scientific novelty of the study lies in substantiating the practical feasibility of hybrid methods and developing an optimized approach for short- and medium-term forecasting of currency trends.

The practical significance of the results is reflected in the potential application of the MNKY method by financial analysts, banking institutions, investment funds, and small and medium-sized enterprises for managing currency risks, developing anti-crisis strategies, and making informed financial decisions.

Additionally, the proposed approach can serve as a foundation for creating intelligent financial monitoring and forecasting systems, enhancing overall efficiency in currency and financial markets.

Keywords: methods, ARIMA, GARCH, exchange rate, forecasting, fundamental analysis, technical analysis, time series, financial markets, volatility, machine learning.

A COMPREHENSIVE REVIEW OF ARTIFICIAL INTELLIGENCE-BASED UAV NAVIGATION METHODS

2025-06-26T06:35:50+00:00

Vehicles (UAVs), delving into the evolution, methodologies, and emerging trends that define this dynamic field. It provides a comprehensive review of state-of-the-art techniques in UAV navigation by categorizing them into two primary paradigms: optimization-based methods and learning-based methods. The article begins with a historical overview that outlines key milestones and technological breakthroughs — from early deterministic, rulebased algorithms to advanced AI-driven systems — laying the foundation for understanding current approaches.

The article describes optimization-based methods, considering both classical and advanced techniques, including algorithms such as Dijkstra, A*, Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Differential Evolution (DE), Simulated Annealing (SA), Genetic Algorithms (GA), Grey Wolf Optimization (GWO), and Pigeon-Inspired Optimization (PIO). It details their fundamentals, operating principles, and the recent modifications that researchers have employed to meet specific navigational objectives. Similarly, the review categorizes learning-based methods by examining Reinforcement Learning (RL), Deep Reinforcement Learning (DRL), Asynchronous Advantage Actor-Critic (A3C), and Deep Learning (DL) techniques, emphasizing various proposed approaches, their benefits, and the goals they aim to achieve.

As a result, this article offers a comprehensive analysis of existing navigation methods, describing their features, drawbacks, and inherent complexity. Although AI-driven navigation can be computationally expensive, the significant improvements in flexibility and overall performance enhance UAV robustness in complex dynamic environments. These findings provide insights for researchers and developers, helping to choose the most suitable approach for their work while highlighting the promise of hybrid strategies that combine the deterministic reliability of optimization techniques with the adaptability of learning-based methods.

Finally, the review identifies current research gaps — such as the need for improved big data processing, increased computing power, enhanced energy efficiency, and better fault handling — and outlines future research directions to accelerate advancements in autonomous UAV navigation. These insights provide clear guidance for future studies aimed at developing more robust, scalable, and efficient UAV navigation systems.

Keywords: navigation, artificial intelligence, unmanned aerial vehicles, UAV, path planning, autonomous navigation, trajectory planning, obstacle avoidance.