Viktor Khavalko

The paper substantiates the factors that give rise to crisis situations and conflicts during the management of energy-active facilities – situations that frequently escalate into emergencies requiring rapid response and effective decision-making. In such contexts, it is crucial to examine the psychological and intellectual resilience of operators, as this resilience underpins the capacity of systems and agents to function within the scope of defined objectives and authorities. Such resilience serves as a key factor in maintaining functional efficiency when countering various types of threats and attacks. A significant aspect of functional resilience involves analysing the logic behind decision-making aimed at achieving system-oriented outcomes. These decisions are grounded in principles of coherence, multi-level analysis, and risk assessment. Identifying the structure of this management logic and the associated decision-making procedures enables the development of efficient algorithms for addressing crisis situations – algorithms applicable across diverse fields, ranging from security management to socio-economic systems. The challenge of making efficient decisions – whether by an intelligent agent (a team player), an operational team, or a managerial/administrative structure – is determined by a complex interplay of psycho-neurophysiological and intellectual factors, operating at both the individual and team levels. They are participants in a game structure that shapes the strategy of permissible actions. To address this problem, the paper develops a cognitive-logical concept for the operational resilience of cyber-physical infrastructure, which integrates a “human-agent” into a unified situational control loop. Based on the category algebra, a category-theoretic information model is constructed for the decomposition of a hierarchical production structure, accounting for 16 threat-resilient personnel levels (P1–P16). By integrating H. Gardner’s theory of multiple intelligences and R. Sternberg’s triarchic theory, a cognitive map for operator decision-making during laser-based environmental monitoring is formalized. A mathematical framework for the quantitative assessment of personnel actions is developed, incorporating cognitive coefficients (KF), confidence levels (K(B)), and error risk factors (Ri). Practical testing of the system at the Burshtyn Power Station (Ukraine) and glass-manufacturing companies has demonstrated its efficiency under extreme operating conditions (1200°C–1500°C). Implementation of this method has reduced the reaction time of operational and dispatch personnel during pre-emergency and emergency conditions (Alarm/Avaria) and minimized the overall risk of error decisions stemming from the human factor.

Traffic flow prediction remains a cornerstone of intelligent transportation systems (ITS), facilitating congestion mitigation, route optimization, and sustainable urban planning. Graph Neural Networks (GNNs) have revolutionized this domain by adeptly modeling the intricate graph-structured nature of traffic networks, where nodes represent sensors or intersections and edges denote spatial relationships. Recent years (2023–2025) have witnessed a surge in scientific innovation, with several novel approaches pushing the boundaries of traffic prediction accuracy and robustness. Notably, hybrid GNN-Transformer architectures have emerged, leveraging the spatial reasoning of GNNs and the temporal sequence modeling power of Transformers to capture long-range dependencies and complex spatiotemporal patterns. Physics-informed GNNs integrate domain knowledge, such as conservation laws and traffic flow theory, directly into the learning process, enhancing interpretability and generalization to unseen scenarios. Uncertainty-aware frameworks, including Bayesian GNNs and ensemble methods, provide probabilistic forecasts, crucial for risk-sensitive applications and adaptive traffic management in volatile urban environments. This article provides a comprehensive guide to implementing GNNs for traffic flow prediction, detailing best practices in data preparation (e.g., graph construction, feature engineering, handling missing data), model training (e.g., loss functions, regularization, hyperparameter tuning), and real-time deployment (e.g., edge computing, latency optimization). We critically compare GNNs to traditional statistical and deep learning methods, highlighting their superior ability to capture non-Euclidean spatial dependencies, adapt to dynamic and evolving network topologies, and seamlessly integrate multi-modal data sources such as weather, events, and sensor readings. Empirical evidence from widely used benchmarks, including PeMS and METR-LA, demonstrates that state-of-the-art GNN models achieve up to 15–20 % improvements in accuracy metrics such as Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) over conventional baselines. These gains are attributed to the models’ capacity for dynamic graph learning, attention-based feature selection, and robust handling of heterogeneous data. Drawing on these recent innovations, this synthesis highlights GNNs' pivotal role in fostering resilient, AI-driven traffic systems for future smart cities, setting the stage for next-generation ITS solutions that are adaptive, interpretable, and scalable. In addition to these advancements, the integration of real-time sensor data and external information sources has further improved the responsiveness of traffic prediction models. Modern GNN frameworks are capable of handling large-scale urban networks, making them suitable for deployment in metropolitan areas with complex road infrastructures. The use of transfer learning and domain adaptation techniques allows models trained in one city to be effectively applied to others, reducing the need for extensive retraining. Furthermore, explainable AI approaches within GNNs are gaining traction, enabling stakeholders to understand and trust model decisions in critical traffic management scenarios. Recent research also explores the fusion of GNNs with reinforcement learning, enabling adaptive control strategies for traffic signals and congestion pricing. The scalability of GNNs ensures that they can process data from thousands of sensors in real time, supporting city-wide traffic optimization. Advances in hardware acceleration, such as GPU and edge computing, have made it feasible to deploy these models in latency-sensitive environments. Collaborative efforts between academia, industry, and government agencies are driving the adoption of GNN-based solutions in smart city initiatives. As urban mobility continues to evolve, the ability of GNNs to incorporate emerging data modalities, such as connected vehicle telemetry and mobile device traces, will be crucial for future developments. The ongoing refinement of model architectures and training protocols promises even greater accuracy and robustness in traffic flow prediction. Ultimately, the convergence of GNNs with other AI technologies is set to transform intelligent transportation systems, paving the way for safer, more efficient, and sustainable urban mobility.

Class imbalance is a prevalent challenge in machine learning, particularly in domains like credit risk assessment, fraud detection, and medical diagnosis, where the minority class often represents the critical cases of interest. Traditional machine learning models tend to be biased towards the majority class, leading to suboptimal performance in detecting minority class instances. This paper explores various strategies to mitigate the effects of class imbalance, including performance metric adjustments, resampling techniques such as oversampling, undersampling, and Synthetic Minority oversampling Technique (SMOTE), as well as algorithmic modifications. We propose different hybrid approaches that combines Random Forest with Laplace regularization and others to enhance the classification of minority classes in imbalanced datasets. Our approach is validated using a credit risk dataset, where the proposed method demonstrates significant improvements in recall and F1 scores compared to baseline models, suggesting its effectiveness in addressing class imbalance in critical applications.

Automation and optimization of recruitment processes are crucial components of company operations. One of the key tasks in recruitment automation is the analysis of resumes, which often have a complex structure, complicating their processing by traditional artificial intelligence methods. This study focuses on the development and evaluation of methods for converting resumes from PDF format to text format for further use in artificial intelligence systems. The paper analyzes the basic problems associated with preserving structural and contextual information during the processing of such documents. The use of the multimodal GPT-4 Vision model for reading and converting text from PDF documents is proposed. A comparison is made between the results obtained using GPT-4 Vision and traditional OCR Tesseract. BLEU, ROUGE-1, and ROUGE-L metrics were used to evaluate accuracy. A total of 157 resumes were collected and analyzed, revealing an increase in discrepancies between texts obtained using GPT-4 Vision and OCR as the volume of text increases.

Very often in everyday life it is necessary to regulate a certain process and maintain a certain state, despite the negative effects on productivity. This is especially true in the medical field, where devices and equipment are increasingly common, on which, after receiving a picture, additional methods of image processing can be applied, before the direct use of boundary detection methods. Most of these devices use simple or complex algorithms to perform tasks. That is why the relevance of studying different aspects of segmentation algorithms, analyzing and comparing different types of medical images, conducting experiments on them, comparing the use of different methods and their settings is unquestionable. In order to conduct experiments and comparative analysis, the article proposes to create a process simulation system that will be regulated by the user. The results made it possible to form recommendations regarding the conditions of specific application of the methods.

The paper deals with the process of building the architecture of an automated control system for the greenhouse operation. The advantages and disadvantages of the traditional greenhouse ACS architecture, which consists of three major components - the greenhouse, the IoT platform and the ML model - are analyzed. In order to avoid the disadvantages of such ACS, it is proposed to use cloud technologies. The greenhouse ACS architecture has been developed based on cloud technology and an IoT platform that is flexible, reliable, mobile and versatile. The results of the operation of the greenhouse, the microclimate of which is controlled and maintained with the help of automatic control sys- tem, which is built on the basis of the developed architecture, are presented.

The problem of images classification is considered. It is proposed to use the autoassociative neural networks based on the model of successive geometric transformations for this problem solving. The advantages and disadvantages of this approach are considered, as well as ways to improve the accuracy of image classification.

The questions of neural networks usage for cryptographic information protection have been considered. The problem of cryptographic stability and the main reasons for the unreliability of cryptosystems have been investigated. The issue of neurocomputers application in cryptanalysis is considered. The analysis of the neurocomputers work based on the developed software is done.

The “model of successive geometric transformations” paradigm has been adapted for the implementation of parallel-streaming neural network encryption-decryption of data in real time. A model and structure of a parallel-streaming neural-like element for the mode have been developed.

Currently optical access networks are developing towards new converged architecture. Network convergence allows to integrate multiple access networks with different purposes such as home area network, enterprise network and mobile backhaul network. In this paper, we propose a new algorithm of wavelengths rearrangement and load balancing (WRLB) based on optical wavelength-time division multiple access (OWTDMA). Proposed algorithm improves the flexibility, scalability and energy efficiency of optical access network by using time-frequency resource blocks for frames transmission.