Computer systems and information technologies

METHOD FOR COMPUTER NETWORK TRAFFIC ANALYSIS BASED ON ENTROPY CHARACTERISTICS AND MULTIVARIATE MATHEMATICAL STATISTICS

Olha ATAMANIUK, Volodymyr DUDNYK, Nadiia LYSENKO — Sun, 31 May 2026 00:00:00 +0300

Modern computer networks generate traffic whose behaviour changes over time not only in volume but also in internal structure. Because of this, anomaly detection cannot be reduced to fixed thresholds on separate metrics; it must account for changes in address, port, and protocol distributions together with the joint variation of interrelated traffic descriptors.

This paper presents a method for computer network traffic analysis based on entropy characteristics and multivariate mathematical statistics. The method transforms packet or flow observations collected within a time window into a state vector that combines entropy measures of categorical traffic attributes with volumetric, dispersion, and flow descriptors.

The proposed approach includes formalization of the traffic analysis process, construction of an informative feature system, a multivariate model of normal traffic states, and a structural model of the detection procedure. Algorithmic implementation is organized as a sequence of window formation, empirical distribution estimation, entropy computation, standardization, principal component transformation, multivariate statistical control, and interpretation of feature contributions.

The paper also outlines a methodology for evaluating the developed method in terms of detection quality, robustness to parameter settings, sensitivity to structural changes, and interpretability of monitoring decisions. The resulting framework is intended for traffic monitoring tasks in which payload-independent analysis and adaptation to non-stationary network behaviour are required.

INTEGRATION OF LASER AND OPTO-GALVANIC METHODS IN ENVIRONMENTAL MONITORING SYSTEMS OF TECHNOGENIC FACILITIES

Liubomyr SIKORA, Nataliia LYSA, Olga FEDEVYCH — Sun, 31 May 2026 00:00:00 +0300

The current stage of development of the electric power, chemical, mechanical engineering, and printing industries is characterized by the extensive use of a wide range of resource components, including coal, oil, natural gas, paints and coatings, and polymers, which are marked by increased environmental aggressiveness. Highly intensive production operating modes, driven by market demands, lead to a significant growth in resource consumption within energy-intensive technological processes. This, in turn, results in increased concentrations of dust as well as harmful gaseous and liquid emissions released into the atmospheric and aquatic environments, causing an overall deterioration of environmental conditions. At the same time, real-time assessment of the environmental state is often complicated by the limited capabilities of data acquisition and processing using conventional measurement methods.

A considerable number of information and measurement systems, laboratory methods, and analytical instruments exist for determining the concentration of harmful emissions in the atmosphere, soils, and water bodies of ecosystems; however, most of these approaches are based on environmental sampling followed by laboratory analysis, which requires a certain amount of time. In technological systems, control and measurement devices and sensors are employed; nevertheless, they do not ensure the measurement of critical production process parameters during the stages in which pollutant flows are generated.

Typical examples include emissions of combustion products in boilers of thermal power plant units and municipal systems, as well as in construction, transport and aviation sectors, oil refining and gas industries, and chemical plants of both public and private ownership.

In this context, the problem of environmental monitoring—namely, the development of systems, structures, and sensors—is technically urgent and socially significant.

Accordingly, the development of information technologies for data acquisition in environmental monitoring of ecological media (atmosphere, water, and soils), based on opto-galvanic pollution control methods and laser remote sensing techniques for measuring the concentration of harmful gas emissions, combustion products, and dust, is highly relevant, as these technologies enable continuous monitoring of pollutant concentrations.

GENERALIZED METHOD FOR MANAGING THE LIFECYCLE OF TERRAFORM INFRASTRUCTURE ACROSS MULTIPLE ENVIRONMENTS

Denys KOLOMYTSKYI, Pavlo REHIDA, Oksana ONYSHKO, Yuliia ILCHYSHYNA — Sun, 31 May 2026 00:00:00 +0300

The article examines the challenge of managing the lifecycle of cloud infrastructure, as described using Terraform, across multiple environments (development, staging, and production). In industrial settings, multi-environment infrastructure management gives rise to a set of interrelated challenges: the absence of formalized procedures for promoting changes between environments with explicit source readiness verification, fragmented compliance checking that fails to distinguish code-level syntax validation from plan-level semantic verification, and configuration drift detection that lacks classification by severity, generating false alerts for expected changes such as dynamic IP addresses and rotated certificates. An analysis of existing approaches to environmental isolation, configuration compliance verification, and drift detection reveals that none of the current methods address the full lifecycle in a unified manner. A generalized method is proposed, based on a formalized two-projection lifecycle model: the horizontal projection describes an eight-stage finite automaton of an individual environment (Init, Author, Validate, Plan, Comply, Approve, Apply, Monitor), while the vertical projection defines a partially ordered environment space with a formalized promotion operation. The method introduces a source readiness precondition requiring the source environment to be in the monitoring stage with empty planning and drift deltas, versioned configuration snapshots ensuring code identity across environments, multi-level compliance verification across code, plan, and state levels using hierarchical policy inheritance, and a three-class drift classification (critical, actionable, informational) with an effective delta mechanism that filters expected changes. Experimental verification on a real multi-environment AWS infrastructure (27 managed resources, 3 environments) using the Scalr platform confirms that the proposed method ensures automated detection of 100% of policy violations before the apply stage (compared to 50% for GitOps CI/CD and 0% for Terraform CLI), reduces false drift alerts to zero, and decreases the number of manual promotion steps to a single approval for the production environment.

DEVELOPMENT OF A HYBRID MODEL «PHYSICS-INFORMED AUTOENCODER WITH SPECTRAL CONSISTENCY»

Yurii BILAK — Sun, 31 May 2026 00:00:00 +0300

The subject matter of the article is the application of hybrid neural network models, constrained by physical laws, to inverse spectroscopic problems, particularly for the reconstruction of physicochemical parameters of materials from their spectral characteristics. The paper proposes a novel architecture – Physics-Informed Autoencoder with Spectral Consistency, which combines the capabilities of deep learning with prior physical knowledge, specifically the Bouguer–Lambert–Beer law for modeling absorption. The goal is to enhance the accuracy, robustness, and interpretability of models solving ill-posed inverse spectroscopic problems, especially under limited availability of experimental data and the presence of noise and spectral distortions. The tasks to be solved include: the development of a hybrid architecture that integrates a physical forward model and a neural residual correction block; the generation of synthetic spectra using physical modeling, spectral augmentation, and noise simulation; the implementation of active learning for the optimization of the training set; numerical optimization of the network configuration; and a comparative analysis with other architectures. The methods used are based on mathematical modeling of spectral responses, convolutional neural networks (CNN), autoencoders, weakly-supervised training, active learning, and performance metrics such as MSE and R². A series of numerical experiments were carried out on both synthetic mixtures and real spectral data of CuSO₄·5H₂O films deposited by photochemical laser irradiation. The results show that the proposed model accurately reconstructs component concentrations and film thicknesses even under noisy and non-ideal conditions. Conclusions. The scientific novelty of the results obtained is as follows: 1) for the first time, a hybrid neural network architecture was developed for approximating inverse spectroscopy problems, which combines the advantages of data-driven methods and physically based models in the form of a Physics-Informed Autoencoder, in which the physical forward model is integrated directly into the architecture and supplemented by an adaptive correction neural network; 2) the method for restoring physicochemical parameters of materials from spectral data was improved by combining physical modeling with neural network compensation of residual discrepancies; 3) a systematic comparison of hybrid physics-informed architectures was further developed, as a result of which the advantage of the developed model over other variations of the Physics-Informed Autoencoder, as well as over modern neural network methods based on CNN+LSTM and CNN+Transformer in terms of restoration accuracy and physical consistency of results, was shown; 4) the developed architecture provides high accuracy (R² ≈ 0.987), resistance to noise and overlapping spectral lines, as well as physical interpretability of the latent space, and active learning allowed to reduce the data volume by 40% without loss of accuracy.

A FORECASTING METHOD BASED ON CLUSTERING THE POLYNOMIAL EXTRAPOLATION SEQUENCE

Yurii TURBAL, Oleksandr KUBAI — Sun, 31 May 2026 00:00:00 +0300

Time series forecasting is an important task in intelligent data analysis, especially under conditions of short samples, local non-stationarity, noise, and increased sensitivity to external disturbances. These properties are characteristic, in particular, of financial time series, where local trends, random fluctuations, and abrupt changes in dynamics may coexist even over small observation intervals. One of the promising approaches to short-term forecasting is polynomial extrapolation. However, the use of polynomials of different orders for the same segment of a series produces a set of alternative forecast values, which complicates the selection of the final forecast.

This paper proposes a short-term forecasting method based on clustering the values of the polynomial prediction sequence. For a local fragment of a time series, a sequence of polynomial forecasts is formed over a range of polynomial orders, after which cluster analysis is applied to this set of values. The densest interval method and the DBSCAN algorithm are used to identify the dominant forecast region, while the final forecast value is defined as the central characteristic of the detected cluster. The efficiency of the proposed approach is compared with the forecasting method based on averaging the polynomial extrapolation sequence.

Experimental studies were carried out on deterministic functions, stochastic sequences, and real intraday stock data for Netflix using the Close parameter. It was found that the polynomial prediction sequence has an internal structure in the form of local extrema, concentration intervals, and distant values, which justifies the feasibility of its clustering. The scientific novelty of the study lies in refining the mechanism for selecting PPS elements by moving from index-based averaging to structural analysis of the spatial grouping of forecast values. The practical significance of the work lies in improving the robustness of short-term forecasting for financial time series.

METHOD OF SOFTWARE IMPLEMENTATION OF INTELLIGENT ALGORITHMS FOR CONTROL OF UNMANNED AERIAL VEHICLES IN HARD REAL-TIME SYSTEMS

Dmytro MEDZATYI, Stepan TANASIICHUK — Sun, 31 May 2026 00:00:00 +0300

The article resolves the current scientific and technical contradiction between the need to increase the accuracy of precision guidance of unmanned aerial vehicles (UAVs) in conditions of intense interference and severe limitations of computing resources of onboard systems. The presented work focuses on the creation of an intellectually robust control architecture capable of ensuring stable functioning in conditions of time latency, noisy navigation data and dynamic uncertainty of the object. The scientific novelty of the research lies in the development and implementation of a recurrent self-evolving neuro-fuzzy network (RSEFNN), integrated into the adaptive switching mode controller (ASMC) circuit for online identification and compensation of non-stationary external disturbances. The key feature of the developed architecture is the combination of high robustness with computational efficiency, which is achieved through mathematical optimization of transcendental functions using the Padé method, which allowed to reduce resource consumption by six times compared to standard implementations. The use of self-evolving structures with a strict restriction on the number of rules guarantees determinism of execution time (WCET), which is critically important for aviation certification of on-board software. An important practical result was the creation of a software simulation bench in the MATLAB environment based on an object-oriented approach and a fixed integration step of the 4th-order Runge-Kutt method, which ensures full reproducibility of numerical experiments. The developed algorithm for deterministic actuation distribution allows to effectively control a UAV with an excess number of engines without using iterative procedures, guaranteeing a constant execution time of operations regardless of the input signals. Experimental validation using the Dryden spectral turbulence model confirmed the high robustness of the hybrid system under conditions of intense stochastic disturbances typical of low altitudes. Statistical profiling showed that the ASMC+RSEFNN method consumes less than 2 KB of RAM and has a runtime margin of more than 33% relative to the critical limit of 2 ms. Compared to deep learning and nonlinear predictive control methods, this approach demonstrates significantly higher computational efficiency, allowing to combine intelligent noise compensation with strict real-time requirements on ARM Cortex-M7 microcontrollers.

TOOLS FOR SEMANTIC SEARCH AND ANSWER GENERATION IN UKRAINIAN-LANGUAGE MUSEUM SYSTEMS

Khrystyna LIPIANINA-HONCHARENKO, Vadym VITENKO, Diana ZAHORODNIA — Sun, 31 May 2026 00:00:00 +0300

This article examines modern semantic search models and a search-based response generation approach for building intelligent museum information systems focused on Ukrainian-language and bilingual textual materials. It is shown that traditional lexical search, based on keyword matching, does not always ensure adequate quality in cultural heritage tasks, where relevance is determined not only by formal term matching but also by the semantic proximity of the query and the document. In this context, text vector representation models take on particular significance, as they allow texts to be represented in vector space and enable the retrieval of contextually relevant information even in the absence of direct lexical matches. This paper analyzes the LaBSE, paraphrase-multilingual-MiniLM-L12-v2, multilingual-E5-large-instruct, and BGE-M3 models in terms of Ukrainian language support, multilingualism, search accuracy, computational resource requirements, and suitability for local deployment. The generation of search-based responses is examined separately as an approach in which a generative model constructs a response based on fragments retrieved by a search module from an external knowledge base. It is argued that this approach is particularly suitable for museum information systems, as it reduces the risk of hallucinations, improves the factual accuracy of responses, and enables a source verification mechanism. The advantages and limitations of open and closed generative models in the context of Ukrainian-language museum services are also systematized. Based on the analysis, practical recommendations are formulated regarding the selection of search models and the configuration of answer generation systems for digital cultural heritage.

IMPLEMENTATION OF INTELLIGENT QUALITY CONTROL SYSTEMS AT FLOUR MILLS IN UKRAINE

Artem HUTSALYUK — Sun, 31 May 2026 00:00:00 +0300

The article is devoted to the urgent problem of modernization of the agro-industrial complex through the transition to automated monitoring of product quality in real time. The author justifies the need to abandon traditional laboratory methods, which have significant time delays (from 2 to 4 hours), which creates risks of producing defective products in the event of technological failures. The proposed solution is based on the development of a cyber-physical system (CPS) based on the Edge Computing architecture. This allows you to transfer the decision-making process directly to the production line, eliminating delays in data transmission to the cloud and eliminating the impact of electromagnetic interference typical of industrial zones. Special attention is paid to safety: the system hardware is designed taking into account the explosive hazard of flour dust (zones 20–22 according to the ATEX classification), which requires the use of sealed housings of the IP65/IP67 standard and limiting the surface temperature of the devices. The technical implementation includes the use of industrial cameras with a global shutter (Global Shutter), which prevent distortion of the image of the moving flour flow. For quality analysis, the MobileNetV2 neural network architecture is used, optimized using TensorRT INT8 quantization, which allows achieving classification accuracy of over 98% with minimal computational costs. The mathematical model of the system is based on the analysis of the CIE Lab* color space and hybrid processing of visual and parametric data from sensors. The scientific novelty of the work lies in the implementation of a dual approach to analysis: in parallel with the classification of varieties, an algorithm based on unsupervised learning (autoencoders) works. This allows you to detect previously unknown types of defects or foreign impurities (insects, metal particles, etc.) by analyzing deviations from the mathematical model of the “ideal product”. The proposed system provides stable operation with a response delay within 15–45 ms, which is critically important for the automatic operation of the defect cutters. The implementation of such a CFS contributes to the harmonization of Ukrainian standards with EU requirements for food safety.

AN ONTOLOGY-DRIVEN KNOWLEDGE-BASED APPROACH TO COMPLEX SYSTEMS MANAGEMENT

Andriy MELNYK, Yurii POPYK — Mon, 01 Jun 2026 00:00:00 +0300

The paper addresses the problem of managing a corporate network as a complex functional system, whose efficiency is determined by the consistency between the infrastructure structure, resource characteristics, and the requirements of application tasks. The feasibility of employing a knowledge-based approach, particularly ontology-based modeling, is substantiated for the formalization of corporate network structure and decision support in its configuration. A conceptual model of a corporate network is proposed, where the network is considered as a multi-level system of interconnected components, including network nodes, services, users, resources, operational parameters, and security constraints. Based on this model, a theoretical foundation for constructing a corporate network ontology is developed, ensuring a consistent representation of infrastructure entities, their properties, and relationships. Furthermore, it enables the formalization of interaction rules and logical constraints in the form of ontological axioms. The study proposes a formalized representation of the corporate network ontology in the form of a tuple-based structure, integrating sets of objects, parameters, states, task performance characteristics, and axiomatic constraints. On this basis, a parameter normalization model is developed, enabling the transformation of heterogeneous characteristics into a unified evaluation scale, taking into account the task context and ontological constraints. This facilitates the construction of an integral configuration quality criterion that incorporates resource, operational, and functional aspects. The main contribution of the work is the development of a method for ontology-driven configuration of a corporate network based on a set of tasks. The method relies on tuple algebra operations to generate a set of feasible configurations that satisfy structural, parametric, and axiomatic constraints, followed by the selection of an optimal configuration using multi-criteria evaluation. The proposed approach ensures transparency, reproducibility, and interpretability of the configuration process. Experimental studies were conducted for two types of corporate networks: a university information and telecommunication system and a network of a commercial enterprise. The results demonstrate that the proposed method significantly improves key performance indicators, including reduced service access latency, decreased computational resource load, fewer access conflicts, and enhanced compliance with security policies. The integral configuration quality criterion decreased by approximately 48–51%, confirming the effectiveness of the proposed approach. The proposed approach can serve as a foundation for the development of intelligent corporate network management systems, as well as for further advancement of optimization methods for complex information and telecommunication systems based on ontology-driven knowledge representation.

METHOD FOR QUANTITATIVE EVALUATION OF THE EMPIRICAL CONFIRMABILITY OF INVARIANT-ORIENTED SIGNALS IN AUTOMATIC SOFTWARE ERROR DETECTION

Frederik HURALNYK, Viacheslav KOVTUN — Sun, 31 May 2026 00:00:00 +0300

This paper addresses the problem of improving the reliability of automatic error detection in software through the integration of formal invariant analysis and machine learning methods. The study focuses on the gap between invariant-oriented formal signals and empirically observed anomalies in program execution, which limits the effectiveness of both formal and data-driven approaches to log and metric analysis.

For the first time, a method for the quantitative evaluation of the empirical confirmability of invariant-oriented signals is substantiated, based on their systematic comparison with operational anomalies in program execution. The proposed method formalizes the limits of applicability of invariant analysis, introduces confirmability as an independent criterion for evaluating the quality of error detection models, and justifies the necessity of integrating formal and machine learning levels within a unified information technology framework.

The method is implemented through the construction of execution transitions as aggregated behavioral units, their multimodal representation based on logs and metrics, and the subsequent alignment of formal and empirical signals within a shared analytical space. Empirical verification was conducted on the LO2 dataset, which represents a microservice environment with execution logs, metrics, and labels of correct and erroneous states.

The proposed approach achieved a harmonic quality measure of 0.854 and a precision–recall area under the curve of 0.873, along with improvements in structural characteristics of the model, including an increase in the consistency coefficient to 0.702 and a reduction in entropy-based mixing to 0.398. It was established that 81.4% of invariant violations have empirical confirmation in execution logs, while 18.6% remain unconfirmed. This quantitatively defines the boundary of effectiveness of formal analysis.

PROTECTED LOCAL ACCESS BASED ON PHYSICALLY-DETERMINISTIC LINKS

Volodymyr KYSIL, Tetiana KYSIL — Sat, 31 May 2025 00:00:00 +0300

This article proposes an approach to organizing protected local access based on physically-deterministic links for intelligent medical diagnostic systems. The relevance of this study is driven by the necessity of ensuring confidential remote access to local computing resources operating within isolated networks or behind restrictive systems, such as NAT. The developed approach is based on the concept of a "blind" registry and out-of-band key transmission via physical media, such as QR codes or their analogs, enabling a zero-knowledge access model. Information interaction is protected by implementing physically-deterministic links that combine temporary tunnels with a mechanism for local metadata decryption in the user's browser using the Web Crypto API.This approach eliminates the possibility of data compromise at the network infrastructure provider level and ensures confidentiality without requiring centralized key management systems (KMS), which is vital for autonomous nodes and ease of use. The practical significance of the method is integrated into the general concept of automated diagnostic and consultation information technology, providing secure remote calls to neural network models deployed on local computers or servers. The proposed solution is the subject of ongoing research aimed at minimizing hardware requirements while maintaining high levels of access simplicity and setup for patient data access

METHOD OF ACCELERATED DATA RECOVERY IN INFORMATION SYSTEM ON MOBILE PLATFORM IN POST-FAILURE MODE

Vitalii TKACHOV, Ihor RUBAN — Sun, 31 May 2026 00:00:00 +0300

The article considers the problem of accelerated data recovery in an information system on a mobile platform (ISMP) in post-failure mode. The relevance of the study stems from the fact that, following an emergency data failure in an ISMP, there is a need to restore the critical part of the data to proper use as quickly as possible, given limited computational, network, and power resources. The aim of the work is to develop a method of accelerated data recovery, which after a data failure provides the selection of the smallest sufficient recovery action based on the control copy and the change log and reduces the duration of the return of the critical data portion to the correct use under current ISMP’s resource constraints. The work formalizes post-failure data recovery through a set of data elements, indicators of a failure, a critical data portion, a set of control copies and a change log. A set of permissible recovery actions is constructed, within which each action is determined by the selection of a control copy, a fragment of the change log, the composition of immediate recovery data and the composition of data for deferred recovery. A rule for selecting the smallest sufficient recovery action and an algorithm for implementing the method have been developed, which takes into account the duration of recovery, the level of residual disruption of the data critical part, the energy resource consumption, and the number of elements for which immediate recovery is provided in the current action. According to the results of the computational study, it was established that the proposed method has an advantage over the basic recovery methods in terms of the recovery duration, the volume of data transmission and reading, the energy resource consumption, the level of residual disruption of the data critical part, and the proportion of successful completion of recovery within the acceptable time interval. The practical significance of the results obtained lies in the suitability of the method for use in onboard and distributed ISMPs, where after an emergency data failure it is necessary to restore the critical part of information base for operation within a short time interval without proceeding to the full recovery of the entire dataset.

ADAPTIVE BIG-DATA MANAGEMENT OF SMART RETAIL ENTERPRISES

Ivan TSMOTS, Volodymyr PETRYNA, Denys RUDAVSKYI — Sat, 31 May 2025 00:00:00 +0300

The paper develops and substantiates an information technology for the adaptive management of a smart retail enterprise based on Big Data. The proposed solution is aimed at creating an integrated closed-loop management environment in which heterogeneous internal and external data streams are transformed into knowledge, forecasts, managerial decisions, and corrective actions in near real time. The study formulates a formalized system of functional, architectural, and operational requirements for such technology. These requirements include adaptability, data integration, scalability, support for batch and streaming processing, low decision latency, data security, fault tolerance, service orientation, and the capability of continuous self-learning. The methodological foundation combines the systems approach, the cybernetic approach, and data-driven management principles with methods of data mining, machine learning, forecasting, optimization, and multicriteria decision making. A formal structure of the technology is proposed as a set of interconnected subsystems for data collection, integration, storage, analytics, forecasting, decision making, implementation, and monitoring. Their interaction is described as a closed transformation cycle that links data acquisition with feedback-based managerial correction. The paper further develops the structural representation of the technology by distinguishing three completed functional stages: data formation, analytical processing with knowledge generation, and managerial decision implementation. A particular contribution of the study is the formal consideration of the relationship between the data-flow intensity and the throughput of the computing infrastructure. This makes it possible to define a real-time operation condition and to explain how overloads, queues, and excessive delays can be prevented when processing high-volume and high-velocity data streams. To assess the performance of the proposed information technology, a system of partial and integral criteria is introduced. The integral multiplicative efficiency criterion jointly takes into account qualitative, temporal, and resource parameters of all stages of the management cycle. In addition, an adaptability criterion is proposed to evaluate the quality of system response, reaction speed, and resource expenditure under changing operating conditions. The obtained results provide a formal basis for designing scalable intelligent management platforms for smart retail enterprises, improving the consistency of information processes, increasing the quality of forecasts and managerial decisions, and supporting proactive enterprise behavior in a dynamic digital economy.

Keywords: adaptive management, smart retail enterprise, Big Data analytics, intelligent decision support systems, machine learning, real-time data processing, data-driven management.

METHODS FOR DIAGNOSIS OF MELANOMA BASED ON DIGITAL IMAGE PROCESSING AND EXPERT SYSTEMS

Eugene FEDOROV, Tetyana UTKINA, Yaroslav KORPAN — Sun, 31 May 2026 00:00:00 +0300

This article proposes a solution to a pressing scientific and applied problem: the development of a method for diagnosing melanoma based on a fuzzy expert system and digital image processing. The method for calculating melanoma features based on digital image processing proposed in the article includes: conversion of a color image into a grayscale image; conversion of a grayscale image into a binary image based on single-level global thresholding using the Otsu threshold; removal of small objects from the binary image using morphological transformation; formation of a binary matrix of image point membership to the object and a grayscale image of the object; determination of the object boundary in the binary image after morphological transformation based on the Kanna method; calculation of the irregularity of the object boundary in the binary image after morphological transformation; determination of the number of colors based on clustering of the gray-scale image object; rotation of the gray-scale image object; calculation of the diameter of the rotated gray-scale image object; verification of asymmetry based on the rotated gray-scale image object. For the diagnosis of melanoma, this work improved a fuzzy expert system for melanoma diagnosis that uses Sugeno’s fuzzy inference algorithm. An experimental study confirmed that the proposed fuzzy expert system achieves a probability of incorrect decisions regarding melanoma diagnosis of 0.02 and a root mean square error of 0.05. The scientific novelty of the study lies in the fact that the proposed fuzzy expert system represents knowledge about melanoma in the form of fuzzy rules that are understandable to humans; it reduces computational complexity, the probability of making an incorrect decision, and the root mean square error. The proposed solution is scalable and suitable for use in intelligent decision-making systems.

MULTI-DRIFT PREDICTIVE MONITORING FOR EVOLVING INFORMATION SYSTEMS

Vasyl LYASHKEVYCH — Sun, 31 May 2026 00:00:00 +0300

This article addresses predictive monitoring of information systems under conditions of multidimensional functional-state evolution. Unlike conventional monitoring approaches focused on isolated anomalies, failures, or statistical deviations in data streams, the proposed approach treats an information system as a multilayer dynamic object influenced by interacting drift processes. The study considers nine drift types relevant to modern software-intensive and cyberinfrastructure environments: configuration, topology, role, policy, architectural, contextual, semantic, goal, and security drift. It is shown that these drifts affect not only current system parameters but also the validity of monitoring, interpretation, and decision-making processes. The current state of the field is analyzed and the literature is shown to remain fragmented across concept drift detection, multivariate change detection, software architecture erosion analysis, ontology evolution, role and policy evolution, context-aware access control, and self-adaptive systems. To address this fragmentation, the paper proposes an integrated predictive monitoring model based on an extended system state vector and a Predictive Drift Index for early identification of hazardous evolution trajectories. The model combines statistical, multivariate, architectural, contextual, semantic, and security-aware perspectives within a unified framework. A validation protocol is proposed, together with a simulation experiment based on controlled injection of isolated and combined drifts into a nine-dimensional system-state representation. The simulation demonstrates that the integrated predictive index reacts more clearly to multi-drift escalation than isolated indicators and supports earlier identification of degraded, vulnerable, anomalous, and critical trajectories. The proposed approach provides a basis for intelligent monitoring of evolving information systems.

EXPERT SYSTEM FOR CONTROLLING OPERATING MODES OF SOLAR PANELS WITH NEURAL NETWORK-BASED OPTIMALITY ASSESSMENT OF DECISIONS

Yehor TSYBULSKIY — Sun, 31 May 2026 00:00:00 +0300

The rapid growth of solar energy utilization necessitates increasing the efficiency of control systems for photovoltaic installations operating under conditions of variable solar irradiance, temperature fluctuations, and component degradation. Modern photovoltaic systems are characterized by nonlinear behavior, stochastic external influences, and dynamic load conditions. Under such circumstances, traditional Maximum Power Point Tracking (MPPT) algorithms, which are typically based on fixed logic and local optimization procedures, do not always ensure optimal system performance, especially in transient and rapidly changing environments.

Existing approaches to photovoltaic system control primarily rely on classical MPPT techniques, rule-based logic, or monitoring-oriented analytical modules. While these methods provide stability and acceptable efficiency under steady-state conditions, they are often limited in adaptability and do not adequately account for complex interdependencies between environmental and electrical parameters. In particular, conventional solutions lack mechanisms for self-learning, dynamic optimality evaluation, and real-time corre ction of control actions, which significantly reduces their effectiveness under uncertainty and nonstationary operating conditions.

A promising direction for overcoming these limitations is the development of cyber-physical control systems that integrate expert knowledge with adaptive data-driven models. In such systems, an expert subsystem generates control decisions based on formalized rules and domain knowledge, while a neural network module evaluates the quality of these decisions and performs their correction based on learned patterns. This hybrid approach enables combining interpretability and structural clarity of expert systems with the adaptability and approximation capabilities of artificial neural networks.

The use of neural networks allows modeling nonlinear relationships between system parameters, approximating complex objective functions, and adapting to changing operating conditions. At the same time, expert systems provide a transparent and logically structured mechanism for initial decision formation, ensuring reliability and compliance with operational constraints. The integration of these components creates conditions for building intelligent control systems capable of maintaining high efficiency, stability, and robustness of photovoltaic installations.

Therefore, the development of an expert system for controlling operating modes of solar panels with neural network-based optimality assessment of decisions represents a modern and relevant scientific and practical task. Such systems have significant potential for improving energy efficiency, reducing losses, and enhancing the adaptability of renewable energy sources within modern cyber-physical infrastructures.

SHIFT PARAMETER ESTIMATION FOR IMPLICIT NEURAL REPRESENTATIONS OF IMAGES

Anna BEDRATIUK — Sun, 31 May 2026 00:00:00 +0300

Неявні нейронні репрезентації моделюють зображення як неперервні функції координат, що дає змогу виконувати обробку сигналів без дискретизації. Водночас перевірка того, чи відрізняються два зображення лише паралельним зсувом, як правило здійснюється методами, які потребують дискретного подання або глобальних агрегувань, що ускладнює їх пряме застосування до неявних моделей і суперечить їхній неперервній природі. Метою роботи є запропонувати та обґрунтувати алгоритм, який визначає, чи пов’язані дві неявні нейронні репрезентації зображення паралельним зсувом, використовуючи лише значення похідних до другого порядку, а також забезпечити оцінювання параметра зсуву. Запропонований підхід ґрунтується на локальній лінеаризації оператора зсуву та використанні аналітично доступних похідних, отриманих за допомогою автоматичного диференціювання в межах неявної моделі. Критерій узгодженості будується на перевірці стабільності оціненого зсуву на множині точок області визначення. Для підвищення надійності локальні оцінки агрегуються з використанням робастних процедур разом із перевірками узгодженості, що зменшують вплив неоднорідних ділянок сигналу. Розроблено критерій виявлення зсувної еквівалентності двох неявних репрезентацій та процедуру оцінювання параметра зсуву. Показано, що запропонований критерій узгоджується з неперервною природою неявних моделей, не потребує декодування у піксельну ґратку та є придатним для застосування до моделей, у яких похідні доступні аналітично. Запропонований підхід має наукову новизну у вигляді тесту зсувної еквівалентності для неявних нейронних репрезентацій та практичну значущість як інструмент для швидкої перевірки узгодженості, валідації та попередньої нормалізації даних у задачах комп’ютерного бачення.

Отримані результати застосовні в сценаріях, де зображення або поля вже представлені як неявні нейронні представлення, також відомі як нейронні поля. Це включає автоматизовану перевірку узгодженості реконструкцій, попереднє вирівнювання або нормалізацію перед подальшою обробкою, контроль правильності об'єднання полів та підготовку даних у завданнях комп'ютерного зору. Рекомендується використовувати метод першого порядку як швидку ініціалізацію, а метод другого порядку як уточнення для більших переміщень або текстурованих областей. Для стабільного використання методу другого порядку слід вибирати гладкі архітектури INR, а якість апроксимації INR повинна бути достатньо високою, що підтверджено метриками MSE та PSNR.

SENTIMENT ANALYSIS OF PUBLIC OPINION REGARDING THE WAR IN UKRAINE BASED ON REDDIT DATA USING NLP AND MACHINE LEARNING METHODS

Roman LYNNYK, Victoria VYSOTSKA, Lyubomyr CHYRUN — Sun, 31 May 2026 00:00:00 +0300

The article examines public opinion among Americans regarding the war in Ukraine by analysing text data from the social platform Reddit. The relevance of the work lies in the significant influence of public sentiment in the United States on the formation of foreign policy and on public support for Ukraine. The purpose of the study is to conduct a comprehensive sentiment analysis of English-language comments of Reddit users using modern methods of natural language processing (NLP) and machine learning. The work uses a dataset from the Kaggle platform that contains millions of comments on the Russian-Ukrainian war. Preprocessing of text data was carried out, including cleaning, tokenisation, lemmatisation and removal of stop words. For tone analysis, both classical approaches (VADER, TextBlob) and modern machine learning models were used, including Logistic Regression, Random Forest, SVM, XGBoost, and Naive Bayes. A hybrid approach to text vectorisation (TF-IDF with Word2Vec) was implemented. The results obtained allow us to determine the distribution of emotional assessments (positive, negative, neutral), identify thematic clusters of discussions and investigate the dynamics of changes in public sentiment over time. A comparative analysis of the effectiveness of the models based on the main quality metrics was conducted. Particular attention was paid to the specifics of Reddit discourse, including sarcasm, irony and political polarisation. The practical value of the study lies in the creation of analytical tools for monitoring public opinion, which can be used in diplomacy, politics, and the media to develop effective communication strategies.

FOUNDATIONAL ABSTRACTIONS FOR CORE ENTITIES AND QUERY MECHANISMS IN EVENT-SOURCED SYSTEMS

Ihor YANKIN, Yurii GUNCHENKO — Sun, 31 May 2026 00:00:00 +0300

Event-sourced systems represent application state as a deterministic function of an immutable event history, so queries operate over histories and derived views rather than over a single current-state snapshot. The absence of shared, implementation-agnostic definitions and semantic contracts makes nominally similar queries non-comparable and hinders portable cost reasoning. The goal of the study is to develop a compact, implementation-agnostic formal foundation for core entities and query mechanisms in event-sourced systems, which enables mechanism-level cost analysis independent of specific technologies. A theoretical methodology based on formalization and mechanism-level analysis is proposed. The study defines a minimal set of core entity abstractions that determine representation and interpretation in event-sourced systems (events, streams, aggregates, projections, snapshots, versions). On this basis, querying is formalized as contract-defined deterministic evaluation over immutable histories and organized into four mechanism groups: reconstruction, temporal, cross-stream, and retroactive replay, each specified through explicit scope and cut rules with declared ordering/merge, correlation, and version-normalization policies. Portable cost envelopes are derived by expressing selection and evaluation costs through selected evidence size and amortized per-event processing, including explicit contributions from normalization and replay shortening by snapshots. The study formalized implementation-agnostic core abstractions and contract-defined query mechanisms for event-sourced systems and derived cost envelopes. A theoretical experiment on a synthesized banking event dataset confirmed internal consistency, replay equivalence, and reproducibility under semantics-preserving transformations. The proposed formalization fixes the semantic degrees of freedom required for reproducible and comparable querying over immutable event histories and provides a reusable basis for mechanism-level cost reasoning across implementations. Further research should extend the framework toward practice-complete semantics by formalizing admissibility rules under distributed time uncertainty or read-model staleness under eventual consistency.

AN INTEGRATED KPI FRAMEWORK FOR ORGANISATIONAL INFORMATION SYSTEMS: MAPPING SEVEN GOVERNANCE - LEVEL METRICS TO ISO/IEC 25010:2023 ATTRIBUTES

Mark ISRAEL, Vyacheslav KHARCHENKO — Sun, 31 May 2026 00:00:00 +0300

Information systems have become strategic organizational assets, but classical success literature in the field of information systems offers qualitative models without predetermining which measurement attributes implement each dimension. This article proposes an integrated framework of key performance indicators (KPIs) that bridges this gap by mapping seven new government - level indicators to a vocabulary of measurement attributes taken from the ISO/IEC 25010:2023 standard. The subject of the study is an integrated framework of key performance indicators (KPIs) for evaluating organizational information systems at the government level. The purpose of the study is the group of organizational information systems along with their dimensions of quality, success, and risk in private and public sector deployments. The aim of the paper is to formulate, justify and formulate an integrated framework that maps seven governance - level metrics: Organization - Wide Effectiveness (OWE), Decision Lag (DL), Artificial Intelligence Adoption (AIA), Management Attention Index (MAI), Executive Engagement Score (EES), Risk Response Compliance (RRC) and Risk Mitigation Response Rate (RMAR) – to 25 measurable quality attributes taken from the ISO/IEC 25010:2023 standard. Mapping or suggesting dependencies is done using 37 literature - based dependency links. To achieve this goal, the paper addresses the following tasks:

To organize classical information systems success theory, technology acceptance models and contemporary key performance indicator (KPI) hierarchies into a single conceptual vocabulary.
To define the seven governance - level metrics in a form that is ready to be measured using variables observable from organizational data.
Build an explicit dependency mapping between the seven metrics and the 25 characteristics of ISO/IEC 25010:2023.
Formulate sector - specific weighting profiles for private and public sector deployments; and specify a three - stage empirical validation plan.

As a result of the study, a literature - based framework is presented that combines classical information systems success theory with contemporary constructs of AI adoption, management engagement, and project risk governance. The framework is implemented using a three - tier architecture, produced for both private and public sector deployments. The framework is illustrated using a use case diagram and a customer journey map, and is based on a life cycle interpretation of risk metrics based on a defect profile. A three - stage empirical validation plan (Delphi panel, case implementation, and cross - sector Partial Least Squares Structural Equation Modeling: PLS - SEM survey) is proposed for further research.

COMPARATIVE ANALYSIS OF MISSING DATA IMPUTATION METHODS IN BIOMEDICAL RESEARCH: IMPACT ON BIOLOGICAL AGE PREDICTION

Volodymyr SLIPCHENKO, Liubov Poliahushko, Oleksandr VOLKOV — Sun, 31 May 2026 00:00:00 +0300

Missing data remain a major challenge in biomedical research because they can bias statistical estimates, reduce predictive accuracy, and compromise the robustness of scientific conclusions. The present study provides a comparative evaluation of five imputation approaches: IterativeImputer with RandomForest, ExtraTrees, and BayesianRidge estimators, together with KNNImputer and median-based SimpleImputer. The methods were assessed on two biomedical datasets, Bones (3,285 records, 11 biomarkers, n/p = 299) and NHANES (11,016 records after reduction from 55,081, 85 biomarkers, n/p = 130), with an n/p gradient ranging from 19 to 299. The experimental design incorporated three missingness mechanisms, MCAR, MAR, and MNAR, and three missingness levels: 10%, 40%, and 80%. Imputation quality was quantified using RMSE, while downstream effects were examined through biological age prediction based on ElasticNet and PCA models. IterativeImputer with ExtraTrees achieved the lowest average RMSE (9.275), whereas BayesianRidge and RandomForest demonstrated the strongest average rank (2.19-2.20), indicating more stable overall performance across heterogeneous scenarios. Under MNAR conditions, RandomForest produced the best results (RMSE 10.896), while ExtraTrees was most effective for MAR (RMSE 8.704). Downstream analysis showed that PCA yielded lower prediction RMSE than ElasticNet (2.14 versus 5.86), although 34% of cases exhibited negative correlations. A paradoxical improvement in imputation quality with increasing missingness was observed in 55-75% of scenarios. Median imputation was the fastest method (0.0075 s), whereas RandomForest was the slowest (261 s). The findings support practical recommendations for selecting imputation strategies according to dataset structure, missingness mechanism, and computational constraints in biomedical applications.

ARTIFICIAL INTELLIGENCE IN FINANCIAL TECHNOLOGY: METHODS, APPLICATIONS, AND CURRENT DEVELOPMENTS

Serhii SAVCHENKO — Sun, 31 May 2026 00:00:00 +0300

This study systematises modern methods of generative artificial intelligence, specifically large language models (LLMs), and analyses approaches for their application in the financial technology sector. It provides a summary of existing strategies for using LLMs in FinTech, including zero-shot, few-shot, fine-tuning, Retrieval-Augmented Generation (RAG), and training models from scratch. A comparative analysis of their cost and implementation complexity was performed, identifying the most suitable LLM integration options depending on the application task. The paper presents an algorithm for automatic investment portfolio rebalancing, which combines classical Markowitz Portfolio Theory (MPT), price forecasting using LSTM networks, and technical analysis signals. An extended version of the rebalancing algorithm is proposed, supplementing traditional quantitative methods with two LLM components: a market sentiment analysis module and a financial statement processing module. Integrating these components enables the processing of unstructured data, such as financial news, social media posts, and quarterly or annual corporate reports. Using such data significantly expands the input datasets for price forecasting models, which can improve the quality of investment decisions. Based on the analysed scientific publications, it is shown that combining technical and fundamental financial indicators with market sentiment assessment helps to increase the accuracy of price forecasting for financial instruments. The paper demonstrates the potential for using the proposed investment portfolio rebalancing method in automated financial advisory systems (Robo-Advisors). The main limitations of the study are highlighted, in particular the need to test the rebalancing algorithm in practice using real market data. Directions for further research are identified, relating to the experimental testing of the proposed model on historical data from various periods and the subsequent optimisation of LLM components based on the results of the experiments.