Работая с сайтом, я даю свое согласие на использование файлов cookie. Это необходимо для нормального функционирования сайта, показа целевой рекламы и анализа трафика. Статистика использования сайта обрабатывается системой Яндекс.Метрика

Принять

Научный журнал Моделирование, оптимизация и информационные технологииThe scientific journal Modeling, Optimization and Information Technology

Online media
issn 2310-6018

submit
a manuscript

metadata of articles for the last 2 years

Method for recovering a biometric key in a trusted execution environment and deriving a local session key for secure operations on a user’s client device

2026. T.14. № 3. id 2125

Volkhonskiy V.V. Kaliberda I.V. Pisarenko E.A. Vasilenko S.R.

DOI: 10.26102/2310-6018/2026.54.3.003

A method is proposed for locally recovering a reproducible biometric secret key within a trusted execution environment using an error-tolerant key-extraction construction, followed by deriving a local session key. The protocol architecture explicitly distinguishes: (I) a shared secure-channel key computed by both parties solely from the outcome of a hybrid authenticated shared-secret establishment procedure with a post-quantum component and the handshake transcript; and (II) a local session key computed only on the client device within the trusted execution environment based on the result of local biometric verification. The local session key is used to protect local artifacts and to perform critical on-device operations; it is neither transmitted to the server nor required for server-side verification. The method ensures reproducibility under intra-class variability of biometric measurements, minimizes server-side handling of biometric-derived key material within the organization’s information system, and provides cryptographically sound separation of key-material domains. The object of study is the external communication channel between the user terminal and the company’s remote server; inter-server links between the company server, the cryptobiometric system, and the remote database of the Unified Biometric System are assumed to be protected using certified cryptographic mechanisms compliant with national standards and are not analyzed.

Keywords: trusted execution environment, biometrics, key recovery, noise-tolerant recovery, key derivation function, hybrid authenticated key establishment, post-quantum shared-secret establishment, secure communication channel, handshake transcript, remote identification

Simulation of orbital dynamics of satellite communication spacecraft by linear interpolation

2025. T.13. № 4. id 2124

Sklyarov V.A. Sklyarov A.V. Кузьменко Р.В. Kalach A.V. Соловьев А.С. Durdenko V.A.

DOI: 10.26102/2310-6018/2025.51.4.059

Space technologies are increasingly permeating our lives. Global positioning systems, communications, geology, hydrology, agriculture, and military affairs are just a few of the industries that rely on satellite data. The Russian Federation's satellite constellation is growing every year, and the satellite control system is correspondingly becoming more complex. The high speed of satellites requires constant recalculation of their coordinates with high accuracy. The SGP4 models for low-orbit satellites and the SDP4 models for high-orbit satellites are most commonly used for this purpose. These models provide sufficient calculation accuracy but require significant computing power. When controlling a large number of objects, the computational load can be excessive. This paper is devoted to assessing the error in interpolation calculations of spacecraft positions in low-Earth orbit. To simplify and accelerate calculations of low-orbit satellite dynamics, this paper proposes a linear interpolation method for accelerating satellite position calculations. The results from the SGP4 model were used as a benchmark. A comparative analysis of the accuracy of calculations using the linear interpolation method and the SGP4 model was conducted. It was determined that the acceptable time interval for interpolation should not exceed 60 seconds, ensuring correct interaction between ground stations and the spacecraft.

Keywords: spacecraft, linear interpolation, calculation error, SGP4, low Earth orbit, directional pattern, time interval, spacecraft location

Model and method for evaluating the weighting coefficients of an ensemble machine learning model in the task of forecasting railway freight rates

2026. T.14. № 1. id 2123

Bukharova K.A.

DOI: 10.26102/2310-6018/2026.52.1.005

This article examines the effectiveness of a developed ensemble machine learning model for forecasting rail freight rates. Russian Railways data for a three-year period, comprising approximately 50 million freight shipment records, serves as the empirical base. This dataset ensures a representative sample and accounts for industry-specific data diversity. An ensemble model is developed using the Random Forest, XGBoost, LightGBM, and CatBoost algorithms, with a meta-level implemented as a multivariate linear regression. The ordinary least squares method and Tikhonov regularization are used to calculate the weighting coefficients. This approach stabilizes the solution and reduces the impact of correlated outputs from the base models. Results of computational experiments have shown that combining heterogeneous models into an ensemble improves forecasting accuracy compared to individual algorithms. The average absolute error decreased by 7–13 %, and the average absolute percentage error by 6–12 %, while the determination coefficient increased to 0.942. Additionally, the ensemble's stability was assessed using a sliding window method, which allowed us to determine forecasting horizons that maintain stable results. An extended analysis of the ensemble's behavior with varying input features showed that the model is robust to moderate data distortions and maintains high calculation reproducibility. The obtained results confirm the practical significance of the proposed approach for transport analytics, transportation planning, and the development of economically sound pricing policies.

Keywords: machine learning, ensemble models, tikhonov regularization, least squares method, model accuracy, model stability, railway transport

Intellectualization of decision-making in the procurement of products using a search agent in multivariate analysis of product attributes

2025. T.13. № 4. id 2121

Gusev P. Samotin I. Danilov A.

DOI: 10.26102/2310-6018/2025.51.4.056

The relevance of this study is determined by the need to develop methods for using significant volumes of text information in making management decisions. Therefore, the article aims to identify the key factors and features of intelligent software agents used to search for and process information from open sources. The factors influencing the adoption of management decisions when allocating resources for the purchase of products are considered. A situation of information search in a limited time period is defined, which leads to making a management decision under conditions of incomplete information about products. Descriptions of information search without the use of an intelligent search agent and with the use of an intelligent search agent are proposed. Requirements for the used search agent are defined, consisting of information search accuracy, information search time limitation, and resource limitation for search agent development. Formalized descriptions of the choice of the used artificial intelligence model, requirements for decision-making time, requirements for accuracy and time spent on text information search, as well as an assessment of the cost indicator for developing a search agent are proposed. The implementation of the proposed search agent model was tested on 100 product names related to numerically controlled equipment. Relationships were identified between the number of known attributes and the accuracy of finding new attributes.

Keywords: intelligent agent, artificial intelligence, decision making, information retrieval, search agent model

Integrated monitoring of transport system performance indicators

2026. T.14. № 1. id 2120

Golovnin O.K. Chekina E.V. Ivanova D.M.

DOI: 10.26102/2310-6018/2026.52.1.003

The article is devoted to the urgent problem of developing a unified information space for continuous monitoring of indicators of the functioning of transport systems to provide information support when making managerial decisions to ensure the effective and safe functioning of the urban environment. The paper considers the task of moving from a disparate analysis of individual components of transport systems to a comprehensive and systematic view of transport infrastructure based on integrated platforms that combine heterogeneous data to solve complex urban mobility problems. A methodology for integrated monitoring of transport performance indicators is proposed, designed to aggregate and analyze data from multiple sources in a single digital space in the process of making strategic management decisions in transport. The methodology provides a consistent transition from a structured description of the data processing procedure to the corresponding information data model. The methodology is implemented as part of a digital integrated monitoring platform. Testing has been performed on available open data reflecting the state of the St. Petersburg transport system. The application of the obtained results makes it possible to quickly identify anomalies, evaluate the effectiveness of transport initiatives and diagnose the condition of the transport system, as well as make informed management decisions. The materials of the article are of practical value for specialists in the field of transport infrastructure and urban mobility management.

Keywords: integrated monitoring, road safety, urban mobility optimization, intelligent transport system, digital platform

Intellectual search for analogous patents based on graph representations of the invention's structure

2026. T.14. № 3. id 2117

Korobkin D.M. Fomenkov S.A. Malkov A.N. Kozina S.A.

DOI: 10.26102/2310-6018/2026.54.3.014

The relevance of this work stems from the fact that traditional patent search systems, which are based on relational databases and keywords, are unable to effectively capture the rich context and complex semantic relationships inherent in patent data. The method of intellectual search for patent-analogs based on subgraph isomorphism in a graph database storing component structures of devices described in inventions is proposed. Intelligence manifests itself in the ability of the system to "understand" the structural essence of the invention, to abstract from the text description and to find technically close solutions even in the case of non-matching keywords. The component structure of the devices was obtained by analyzing patent texts using a previously developed neural network model. A patent is represented as a graph, where nodes represent the elements of the invention and edges represent their relationships, enabling the use of graph algorithms to identify relevant patents. Algorithms have been developed for: parsing a JSON file describing the component structure and loading the information into a graph database; comparing graph representations of the invention's component structure; and editing the graph representation of the invention's component structure. The practical significance lies in the developed patent similarity search module, which is based on graph representations of an invention's component structure. This module can be useful during the stages of filing a patent application by the applicant and conducting a patent examination by a patent office examiner. The software module is implemented in Python using the Neo4j graph DBMS.

Keywords: patents, graph DBMS, invention component structure, graph comparison, neo4j

Investigation of the noise immunity of signal-code constructions in various channel models

2025. T.13. № 4. id 2116

Vu T. Glushankov E.

DOI: 10.26102/2310-6018/2025.51.4.061

The increasing demands for throughput and reliability in modern wireless communication systems necessitate effective methods of error-control coding and multi-level modulation. Signal-code constructions, particularly trellis-coded modulation and bit-interleaved coded modulation, are fundamental approaches for solving this task. This paper presents an analysis of the performance and quality of the specified constructions, carried out through simulation in the MATLAB environment. The study covers a wide spectrum of channel models: from the additive white Gaussian noise channel to channels with Rayleigh and Rician fading, featuring both flat and frequency-selective characteristics. In the context of analyzing signal-code constructions of the bit-interleaved coded modulation type, the role and quantitative effectiveness of various interleaver types are investigated in detail. The obtained results not only define the performance limits of each scheme but also form a basis for the future development of methods for their improvement. The simulation results show the high efficiency of both types of signal-code constructions in the flat Rician channel and, furthermore, confirm the key role and performance differences of interleavers in enhancing overall noise immunity under channel fading conditions.

Keywords: signal-code constructions, trellis-coded modulation, bit-interleaved coded modulation, interleaver, MATLAB

Mivar model and algorithm of information processing for avoiding dynamic obstacles with a changing safety zone

2026. T.14. № 1. id 2115

Shen Q.

DOI: 10.26102/2310-6018/2026.52.1.004

We propose a mivar model and a mivar algorithm for dynamic trajectory planning in large indoor spaces with dense crowds of moving people. The main idea is to calculate the parameters of an elliptical changing safety zone for dynamic obstacles by integrating semantic object detection and geometric mapping. The mivar model relies on semantic information: object class, speed, and scene factors. The mivar expert system calculates key parameters of the safety zone and prediction position, providing differentiated safety margins for various semantic targets, such as children, the elderly, and adults. The calculated safety zone will be considered in path planning as the size of the obstacle itself. The scientific innovation lies in the use of dynamic ellipses, changing their size depending on semantic information, to determine the safety zone of dynamic obstacles. The mivar algorithm was verified on a platform implemented in Python using Pygame, with integration of the data exchange interface with the Wi!Mi. This allowed for visualization of the planning process, confirmation of effectiveness, and quantification of safety gains. Prospects for research include deep integration with machine learning methods to enhance the robustness and flexibility of rule updates.

Keywords: dynamic obstacles, changing safety zone, semantic perception, mivar expert system, KESMI, robot trajectory planning, safety control, scene recognition, logical artificial intelligence

Methods of combinatorial optimization of taxonomic filters for information processing in financial market forecasting

2025. T.13. № 4. id 2112

Musin I.R.

DOI: 10.26102/2310-6018/2025.51.4.063

The article is devoted to the study of the system analysis of the predictive ability of the tonality of information flows in the cryptocurrency market. A method of system analysis and combinatorial optimization of taxonomic filters for processing news information is proposed to maximize the effectiveness of the tonality coefficient in predicting the dynamics of cryptocurrency prices, taking into account time lags. A weighted tonality coefficient with a logarithmic multiplier of the information flow volume has been developed, accounting for sentiment polarity, event importance level, and news flow intensity. The paradox of the impact level has been experimentally established, in which low-visibility information demonstrates increased predictive ability compared to official high-impact messages due to the effects of information asymmetry and preliminary integration of critical events into prices by institutional participants. Systematic combinatorial optimization of 39 combinations of taxonomic filters for 10 cryptocurrencies revealed the lack of a universal approach to filtering and identified four different patterns of asset response to the information background. For Bitcoin, a correlation of 0.3611 was achieved with a leading lag of +3 days when using a low-visibility information filter, which provides a significant 32 % improvement over the basic method (correlation of 0.2737, lagging lag of −6 days). The method was validated on a corpus of 108637 classified information units for the period June-September 2025 using language models for multi-taxonomic classification.

Keywords: system analysis, time lag analysis, information processing, tonality analysis, correlation analysis, cryptocurrency markets

Algorithm for searching for the exact contour of cell nuclei in images of buccal epithlial cells

2026. T.14. № 1. id 2111

Limanovskaya O.V. Gavrilov I.V. Meshchaninov V.N. Manakova N.S. Reshetnikov E.D.

DOI: 10.26102/2310-6018/2026.52.1.006

The article develops an algorithm for searching for the exact contour of cell nuclei in images of buccal epithelial cells stained by Giemse. The images of the cells were obtained using a Carl Zeiss Primo Star 415500-011 microscope with 1000x magnification. The algorithm is developed in Python 3.11 using the OpenCV computer vision library. The algorithm consists of two stages. The first stage involves searching for the boundaries of the cell nucleus, selecting the nucleus into a rectangular area and saving it to a separate jpg file. At the second stage, the boundaries of the cell nucleus are clarified in the saved file with the cell nucleus and a mask is applied to the image of the cell nucleus along the specified contours. Images of cell nuclei with refined contours are saved in separate jpg files. The obtained images of cell nuclei with a mask applied can be used to study the morphology of the cell nucleus, including in order to search for markers of the rate of aging. The algorithm can be used not only to obtain files with images of the cell nucleus, but can also be integrated into an algorithm for analyzing the morphology of the cell nucleus, used to search for new markers, for example, the process of age-related cell involution and the body as a whole in gerontological research.

Keywords: computer vision, openCV, search for cell nucleus contours, image segmentation, search for contours in an image, gerontology

Computer simulation of sound fields in rock samples with inclusions of pseudofractal structure

2025. T.13. № 4. id 2109

Krizsky V.N. Belyaev V.V. Adamovich O.O. Gorelik M.A.

DOI: 10.26102/2310-6018/2025.51.4.049

This paper examines mathematical and computer models and methods for studying rock samples containing pseudofractal inclusions and the potential for using computer modeling to investigate their structure. Such structures are found, for example, in fractured rock massifs. These structures, in turn, can serve as hydrocarbon reservoirs, and the study and investigation of their properties makes this research relevant. The article proposes a method for computer modeling pseudofractal inclusions in rock samples in the form of parallelepipeds and ellipsoidal dimples, with anisotropic elastic parameters. A calculation method based on volumetric integral representations and integral field equations in the frequency domain is used to mathematically model sound wave propagation in a geological rock sample saturated with anisotropic fractal inclusions (microanisotropy). The article presents the results of computational experiments on a comprehensive study of fractured reservoirs, with the possibility of expanding the study area and modeling the propagation of various types of fields within it. The materials of the article are of practical value for the study of fractured rocks at great depths by applying the obtained algorithm for generating anisotropic pseudofractal inclusions and calculating seismic fields in their presence.

Keywords: pseudofractal anisotropic inclusions, fracturing, computer simulation, seismic field, volumetric integral representations

Method for registering multiphase CT images with an adaptive weight function

2025. T.13. № 4. id 2107

Kulikov A.A.

DOI: 10.26102/2310-6018/2025.51.4.046

The article proposes an improved method for recording multiphase CT scans of the liver, based on a modified displacement formula that combines global affinity transformation, a local B-spline deformation model, adaptive correction based on an intensity gradient and a noise component to increase robustness. The key element is the weight function E(x), which takes into account local differences in tissue density, it limits deformations in dense structures and enhances them in soft tissues and pathologies. The algorithm is implemented in two consecutive stages – first affine, then B-spline registration – using an extended liver mask and deferred cropping of images, which significantly improves convergence and accuracy. Experiments on a set of clinical cases have shown the superiority of the proposed approach over standard methods. A particularly significant increase was achieved for small pathologies (1–1000 voxels), the average DICE coefficient increased from 0.5737 (affinity registration) to 0.6277. The method demonstrates high resistance to artifacts caused by the patient's breathing, noise and contrast inhomogeneity, and also ensures accurate alignment at the boundaries of objects. The results confirm the clinical applicability of the approach for diagnosis, analysis of liver disease dynamics and treatment planning.

Keywords: medical image registration, multiphase CT imaging, liver, affine transformation, b-spline deformation, adaptive correction, noise component, weight function, DICE coefficient, pathology alignment

A Hyperledger Fabric-based digital diploma management system

2025. T.13. № 4. id 2106

Agamirov L. Agamirov V. Toutova N. Bazunov S. Kozhevnikov K. Shchelkunov M.

DOI: 10.26102/2310-6018/2025.51.4.053

The relevance of the study is due to the global problem of falsification of educational documents, which causes significant damage to employers, educational institutions, and society as a whole. In this regard, the article aims to develop and analyze a decentralized system for verification and storage of diplomas based on the Hyperledger Fabric blockchain platform. A distributed system architecture was designed with the participation of key stakeholders: universities, employers, and the Ministry of Education. A system prototype has been implemented, including smart contracts for issuing, verifying, and revoking diplomas, and endorsement policies and private channels have been configured to ensure confidentiality. A comparison of the proposed Hyperledger Fabric-based system with traditional centralized systems and public blockchains is provided. The results showed that the proposed solution ensures data immutability, instant verification of authenticity, and eliminates the need for a centralized trusted intermediary. Thus, the use of Hyperledger Fabric opens up prospects for creating transparent, secure, and efficient national and international systems for managing educational documents. The virtual machines are publicly available. The materials of the article are of practical value to specialists in the field of blockchain technologies, digitalization of education, and human resource management.

Keywords: блокчейн, hyperledger Fabric, распределенный реестр, цифровые дипломы, верификация документов, смарт-контракты, децентрализованные системы, конфиденциальность данных, доверенная сеть

Optimization of computations in the modified Kalman filter for image processing with autocorrelated noise: a comparative analysis of methods

2025. T.13. № 4. id 2102

Osipenko I.N.

DOI: 10.26102/2310-6018/2025.51.4.051

This paper addresses the problem of optimizing computational costs in the modified Kalman filter used for suppressing autocorrelated noise in digital images. Such noise is a common feature of many practical applications, including medical imaging, Earth remote sensing, and real-time video processing. The classical discrete Kalman filter, while being optimal in terms of minimizing the mean square error, is insufficiently effective under autocorrelated disturbances. In this case, a solution to the discrete filtering problem can be obtained by extending the state vector through the inclusion of additional variables that describe the structure of the noise component. This approach enables more accurate signal restoration but leads to a sharp increase in computational complexity due to the growth of matrix dimensionality and memory requirements. To reduce these computational costs, three optimization strategies for optimizing the filtering algorithm are analyzed: the use of sparse matrix representations, which significantly reduce the number of operations for data storage and processing; multithreaded processing on CPUs to increase computational parallelism; and the transfer of computationally intensive procedures to graphics processing units (GPUs). The experimental study involves testing the developed algorithms on the CIFAR-10 dataset, to which artificially generated autocorrelated noise was added. The results demonstrate that the greatest performance gain is achieved with the GPU-based implementation (a 6–7× speedup compared to the baseline scheme), while the effectiveness of multithreading and sparse matrices depends on the dataset size and structure. The findings confirm the potential of the proposed solutions for practical use in high-performance image filtering systems and their integration into modern machine learning methods.

Keywords: kalman filter, modified Kalman filter, autocorrelated noise, image processing, computational optimization, sparse matrices, parallel computing, GPU acceleration, CIFAR-10, machine learning

Biomechanical models for studying collateral blood flow and tension in deep veins occlusion of the lower extremities

2025. T.13. № 4. id 2101

Komarova L.N. Tikhonenkova O.V. Nikolaeva E.A. Komarova U.V. Martynov V.G.

DOI: 10.26102/2310-6018/2025.51.4.052

The relevance of the study is due to the high prevalence of venous thromboembolic complications worldwide and the need for an interdisciplinary in-depth study and improvement of the approach to diagnostic, therapeutic, and preventive measures for acute deep vein thrombosis. There are relatively few scientific studies on collateral blood flow, which is an important mechanism for compensating for venous circulation disorders. In this regard, a computer modeling technique was applied in the SolidWorks environment, which allowed us to study the collateral blood flow in different veins of the lower extremities and the level of tension in the popliteal vein and before the complete occlusion of the common femoral vein in a patient who had suffered acute ileofemoral thrombosis. The work was based on computer tomograms and angiograms, which were interpreted using the Micro Dicom program. Computer biomechanical 3D models of veins with thrombotic masses are constructed according to tomogram slices in the SolidWorks program. The analysis of flow pressure and tension in deep and subcutaneous veins of the lower extremities is carried out. The article shows that collateral veins perform the function of alternative paths for venous blood outflow, their lumen and capacity can increase in conditions of hemodynamic disorders. The greatest effect on the popliteal vein and the common femoral vein is caused by thrombosis of the small saphenous vein due to the peculiarities of anatomy and a smaller number of collaterals, the tension increases by 2.3 and 1.6 times, respectively. The materials of the article are of practical value for vascular surgeons, phlebologists, cardiac surgeons.

Keywords: acute thrombosis, biomechanics, collateral blood flow, occlusion, vein pressure, vein tension, post-thrombotic disease

Management of organizational systems based on spatially-oriented decision support systems

2025. T.13. № 4. id 2100

Yamashkin S.A.

DOI: 10.26102/2310-6018/2025.51.4.055

The relevance of this research is determined by the growing complexity of organizational systems and the need to consider spatial dependencies in resource planning, risk management, and process coordination. Therefore, this article aims to identify the principles and tools for managing organizational systems based on spatially oriented decision support systems capable of integrating heterogeneous data, modeling scenarios, and ensuring the transparency of management actions. The leading approach to the study is the integration of spatial analysis methods, simulation modeling, and multi-criteria assessment, which allows for a comprehensive consideration of the relationship between territorial factors and indicators of efficiency, sustainability, and costs. The empirical part includes the development of a solution architecture, the formation of a bank of spatial layers, the construction of performance indicators, and comparative testing of management decision options on a set of real-world cases. The article presents algorithms for selecting priorities, discloses the rules for constructing decision maps, identifies the effects of taking spatial constraints into account, substantiates a reduction in uncertainty and an increase in the coordination of departmental actions. The results obtained demonstrate improved forecast accuracy, a reduction in coordination time, and increased economic efficiency in resource allocation. The article's materials are of practical value to executives and analysts involved in strategic and operational management, as well as to developers of applied solutions in the field of spatial data analysis.

Keywords: organizational systems management, geographic information systems, spatial data, management performance indicators, risk analysis

Modeling the electrophysical properties of a reflective screen based on fractal nanocomposites for Arctic mobile systems

2025. T.13. № 4. id 2098

Serdechnyy D.V. Korchagin S.A. Andriyanov N.A.

DOI: 10.26102/2310-6018/2025.51.4.041

The electrophysical properties of a reflective screen were modeled. The grid elements have a layered, hierarchically constructed structure based on a nanocomposite, similar to fractal formations. The study aims to address the problem of creating effective, lightweight, and extreme-condition-resistant reflective structures for Arctic mobile systems, where traditional screens face limitations due to their large mass, high aerodynamic drag, and performance degradation during icing. Using the developed specialized software package, the influence of the grid element fractality level on the reflectivity over a wide frequency range was studied in low-temperature conditions and the complex Arctic radio wave environment. It was found that increasing the fractality level significantly expands the operating frequency range compared to traditional periodic structures and improves energy efficiency. The multiresonant nature of fractal geometry provides flexibility in managing spectral characteristics. The results obtained can be used in the development of broadband radar systems, remote monitoring systems for mobile objects, including ships and ground transport, in the Arctic, as well as for the creation of secure communication channels and intelligent information security systems in polar night conditions and ionospheric disturbances.

Keywords: modeling of electrophysical properties, reflecting the screen, nanocomposites, arctic mobile complexes, radar, fractal structure, software complex

Developing a hybrid deep learning model for workload prediction in virtualized data centers

2025. T.13. № 4. id 2097

Martynenkov B.V.

DOI: 10.26102/2310-6018/2025.51.4.054

The relevance of this research stems from the need to proactively manage the workload of data centers based on virtual machine technologies and application containerization. This task requires analyzing historical workload data consolidated as time series for computing resources used over a given period of time, such as CPU and RAM pool utilization and stored by the resource monitoring subsystem of the data center administration service. Therefore, this article aims to explore machine learning methods and technologies that support time series forecasting. The article summarizes the features of machine learning models based on statistical approaches and deep learning principles. It examines the structural and functional components of neural network variants specialized in analyzing time series dependencies and solving forecasting problems. The proposed solution is a hybrid deep learning system based on the sequential application of cascades of one-dimensional convolutional neural networks and bidirectional long short-term memory networks. Approaches to the selection of their structural and parametric characteristics are proposed. The results of a comparative experimental evaluation of the proposed solution with the implementation of a workload prediction system, based on statistical prediction methods are presented.

Keywords: virtualized data center, workload, time series forecasting, machine learning, deep learning, one-dimensional convolutional neural networks, bidirectional long short-term memory networks

Analysis of the impact of time delays on the operation of the chassis control system of an unmanned aerial vehicle

2025. T.13. № 4. id 2096

Khasanov I. Berezhansky N. Akaemov P.

DOI: 10.26102/2310-6018/2025.51.4.044

The relevance of this research stems from the need to improve the reliability and safety of unmanned aerial vehicle landing gear control systems operating under conditions of time delays in control signal transmission. Such delays, which occur during the measurement, processing, and execution of commands, significantly impact the system's response time and can lead to failures during landing. Therefore, this study aims to model the behavior of the electronic component of an unmanned aerial vehicle landing gear control system, taking into account the cumulative time delays that determine the overall response time. The primary research method is numerical simulation in the time domain, performed in the Simintech software environment, which enabled the implementation of a structural block model of signal transmission from the position sensor to the actuator. The results showed that the total system delay is approximately two seconds for an acceptable body tilt angle of no more than fifteen degrees. If this value is exceeded, the control signal is not generated, providing additional protection against emergency conditions. The obtained results confirm the need to optimize the control system architecture and select faster-responding components. The research materials are of practical value for the design and analysis of control systems for aircraft-type unmanned aerial vehicles.

Keywords: unmanned aerial vehicle, landing gear control system, signal time delays, dynamic systems modeling, electronic part of the control system, software environment

Method of intelligent control for directing a light beam onto dynamic objects in stage space

2025. T.13. № 4. id 2095

Teryaev L.N.

DOI: 10.26102/2310-6018/2025.51.4.048

This paper presents an innovative method for the intelligent control of light beam targeting on dynamic objects in stage space. The developed approach overcomes the limitations of traditional manual and hardware-dependent tracking systems by integrating advanced technologies, including computer vision, machine learning, and the concept of a digital twin of the scene. A key aspect of the method is the creation of a unified coordinate system that links the physical space with its digital copy, which significantly improves targeting accuracy, minimizes time delays, and enables centralized group control of multiple lighting fixtures. A Kalman filter is used to predict object trajectories in real-time, while homography matrices are employed for precise coordinate transformation between the vision systems and the robotic fixtures. Testing of the method in a real-world stage environment confirmed its high effectiveness. The average targeting error was no more than 15 cm with a maximum deviation of 29 cm (3 % of the test area scale), and stable tracking of objects moving at speeds of up to 2 m/s was demonstrated. The method shows not only high accuracy and scalability but also significant potential for collecting and representing data on scene dynamics in a digital model for analytics and decision-making.

Keywords: intelligent control, stage lighting, computer vision, digital twin, kalman filter, homography, DMX, object tracking, lighting control automation

Analysis of methods for improving the convergence of triplet loss in multi-label classification

2025. T.13. № 4. id 2094

Petrova I.S.

DOI: 10.26102/2310-6018/2025.51.4.039

The paper focuses on improving the performance of multi-label image classification by enhancing the convergence of the triplet loss function with a flexible margin. To achieve this, several modifications to the loss function itself and the training process were analyzed, aiming to stabilize gradient descent and reach the loss function minimum more efficiently. Based on the analysis, several promising hypotheses were identified for experimental evaluation: adopting the distance computation approach from the focal loss, replacing the linear increase of the margin parameter with a logarithmic one, balancing classes within a batch, and adjusting batch size. The proposed convergence improvement methods were tested on the open-source CIFAR dataset with hierarchical labels. The effectiveness of the selected methods was confirmed: each modification increased the final model accuracy by 2–4%, while applying all of them together improved classification accuracy by 10% for lower-level labels and 12% for higher-level labels. The proposed methods were also evaluated for robustness to dataset noise. It was shown that the convergence improvement method based on batch class balancing is sensitive to data errors and is therefore not recommended for use with noisy datasets.

Keywords: multi-label classification, computer vision, loss function, triplets, metric learning, optimization

Optimization of management in an organizational system with territorial distribution of the results of the activities of objects

2025. T.13. № 4. id 2093

Boklashov I.I. Ivanov D.V. Yakov L.E.

DOI: 10.26102/2310-6018/2025.51.4.057

This article explores the use of an expert-optimization approach to improve management efficiency in organizational systems with geographically dispersed outcomes. The core management mechanism involves the central authority allocating integrated resources among objects within a specific operational timeframe. The study identifies key factors that influence resource allocation priorities to achieve targeted outcomes for these objects. A method is proposed to derive quantitative priority coefficients, based on ranking each component and calculating an additive combination of their values. This method integrates multiple approaches, combining group expert evaluations of priority rules based on temporal characteristics of numerical rankings with individual expertise to determine weighting factors using logical ordering techniques. The article presents the results of logically ordering components that characterize the diverse environments of objects, focusing on resource allocation to meet planned volume within set deadlines. The study emphasizes the importance of using priority coefficient evaluations for optimizing the selection of objects to receive allocated resources. A multi-option optimization model is developed to select objects with the highest priority while satisfying balance constraints. To implement this multi-method approach, numerical optimization is recommended using various search algorithms. Additionally, a group expert evaluation process, led by a primary expert, is proposed to finalize decisions on managing integrated resource allocation in organizational systems with geographically distributed outcomes.

Keywords: organizational system, management, expert evaluation, territorial distribution, optimization, multi-method approach

Evaluating the effectiveness of cyber‑threat monitoring and response centers: limits of time‑based metrics and operational quality indicators

2025. T.13. № 4. id 2092

Pakhomov V.V.

DOI: 10.26102/2310-6018/2025.51.4.040

The paper examines practices for evaluating the effectiveness of cyber‑threat monitoring and response centers under growing telemetry volumes and increasingly complex attacks. It is shown that commonly used indicators such as mean time to detect and mean time to respond mainly capture speed while failing to assess whether available data are sufficient for sound conclusions, whether investigation context is present, and whether investigation steps are reproducible. The study compares international guidance and landscape reports with Russian regulatory requirements and analyzes industry publications. As a result, data sources used by monitoring centers are systematized, typical bottlenecks in the data value chain are identified, and limitations of classic time‑based metrics are highlighted. A simple three‑axis comparison framework is proposed: speed, context, and process. The contribution introduces three computable indicators: context completeness (share of incidents corroborated by at least three independent sources), investigation reproducibility (share of steps executed via approved playbooks with machine‑readable logging), and resilience to peak loads (comparison of service‑level target adherence in peak versus baseline periods), together with an integral manageability index combining speed, accuracy, and completeness. The practical value lies in the feasibility of calculating these indicators using existing security event management systems and incorporating them into monitoring dashboards for audit, resource planning, and cross‑team comparability.

Keywords: SOC, effectiveness assessment, MTTD, MTTR, context completeness, investigation reproducibility, load resilience, SIEM, SOAR, XDR

An adaptive R-Tree-based geosearch algorithm for moving users

2025. T.13. № 4. id 2090

Dorokhin V.A.

DOI: 10.26102/2310-6018/2025.51.4.047

This paper presents an adaptive R*-tree-based geospatial search algorithm tailored for real-time applications involving dynamically moving users. Conventional methods that rely on a fixed point and a circular or square search area neglect user kinematics – specifically speed and heading – resulting in low result relevance, excessive data transmission, and increased load on client devices and network infrastructure. To address this, we propose a modified approach that dynamically shapes an asymmetric search region extended along the user’s motion vector. The size and geometry of this region are adjusted in real time based on current speed and a configurable speed-influence coefficient. Additionally, a "visibility zone" parameter is introduced for target objects, enabling indexing not only of their coordinates but also of their potential display radii, thereby improving filtering accuracy. The algorithm is implemented within a metaspace object routing task and employs an R*-tree with optimized bulk-loading to minimize overlaps of Minimum Bounding Rectangles (MBR). Simulation results demonstrate improvement in search relevance up to 7 times compared to classical radial search, without increasing computational complexity and while reducing network payload. The method is applicable to navigation, augmented/mixed reality, autonomous systems, and other domains requiring real-time spatial relevance.

Keywords: geospatial search, r-tree, spatial indexing, real-time algorithms, search relevance, metaspace

Structuring a digitalized organizational system for customer service and management process

2025. T.13. № 4. id 2084

Bakulin A.Y.

DOI: 10.26102/2310-6018/2025.51.4.045

The article discusses the research results leading up to the implementation of the management decision-making process using optimization modeling. These studies are aimed at forming the structure of models for the functioning of a digitalized organizational system for servicing consumer orders in physical and information environments and its management. The structuring of the functioning of the organizational system in the physical environment was carried out in order to establish the characteristics of the customer order service process corresponding to the relations of the management center with the facilities both at the level of established requirements and at the level of monitoring assessment. Structural modeling of the functioning of the information environment for customer service is associated with the analysis of a digitized organizational system as a human-machine system, which makes it possible to characterize the relationship of the control center with objects, indicators and parameters of interaction between automated devices (non-ergatic elements) and personnel (ergatic elements). The construction of a structural model of the management process in the physical environment of an organizational system is determined by combining the decision-making mode at the expert level with the use of intellectual support modules. At the same time, the solutions are aimed at choosing a topology for combining automated service devices into facilities, a pricing policy option, a resource allocation option, and system development. In the first tasks, intellectual support is provided by embedding a multi-alternative optimization procedure in the simulation of an organizational system. The last task is related to optimizing the allocation of development resources based on a predictive model obtained by machine learning using retrospective monitoring data. In cases of structural modeling of the process of managing the information environment for customer service, intelligent support is based on parametric optimization of balancing the number of non-ergatic and ergatic elements of each object in order to ensure extreme and marginal performance requirements.

Keywords: organizational system, structural modeling, management, intellectual support, decision making, optimization

Model of mivar problem-oriented management and decision-making system in state space

2025. T.13. № 4. id 2083

Aladin D.V.

DOI: 10.26102/2310-6018/2025.51.4.023

The article discusses the development of a mivar expert system for solving management problems in a state space. the study combines the mivar and metagraph approaches to describe the architecture of a system capable of operating with evolutionary knowledge bases and processing information through linear logic and computational processing. A model for a problem-oriented management and decision-making system based on mivar agents interacting with metagraph agents within a hybrid intelligent information system is presented. The features of representing the mivar network through metagraph are described, making it possible to divide knowledge and formalize the interfaces for interaction between different information systems. The developed model is compatible with various approaches to constructing conceptual models of subject areas for solving problems through decomposition into sub-tasks. Particular attention is given to the formalization of cyclicity in the functioning of mivar expert systems, which makes it possible to organize decision-making processes in accordance with calendar or production cycles. The practical significance of this work lies in creating a tool for designing and analyzing mivar expert systems, which helps to improve the architecture of these systems and increase the expressiveness of conceptual models of subject areas when solving management problems in a state space.

Keywords: mivar approach, metagraph approach, mivar network, mivar expert system, hybrid intelligent information system, state space, mivar agent, metagraph agent

Quantization of outlier free quantizable language models

2026. T.14. № 4. id 2082

Khan S. Kabir A. Lukmanov R.A.

DOI: 10.26102/2310-6018/2026.55.4.005

As deep learning models including the LLMs become a part of our daily lives, they continue to require more and more computational cost. The heavy models need a lot of processing power to train and even to make inferences. However, we can reduce this cost by compression techniques such as quantization. Standard quantization of some transformer models comes at the risk of presence of outliers that result in inaccurate results. In this study, we develop a hybrid model which involves using clipped softmax in attention heads of the model during training to mitigate outliers and then applying activations aware weights only quantization on trained model which helps in reducing quantization error by scaling the weights before quantization. We show that our approach results in better handling of outliers, hinted by reduced kurtosis in clipped softmax trained quantized models as compared to vanilla trained quantized models. Overall, our hybrid method not only achieves the best final model performance but does so by effectively suppressing outliers by a factor of 5–7x across key metrics, making the model far more robust to the quantization process.

Keywords: quantization, outlier, perplexity, attention, softmax, kurtosis

Application of the project-oriented method for the formation of competencies in the field of artificial intelligence

2025. T.13. № 4. id 2079

Smirnov D.Y.

DOI: 10.26102/2310-6018/2025.51.4.030

The purpose of the article is to solve the problem of increasing the literacy of the population in the field of artificial intelligence technologies. The author has developed and put into practice a program for teaching citizens AI literacy at advanced training courses in higher education institutions. The developed program is based on two methods that take into account the specifics of the contingent of students who receive knowledge within the framework of the additional education system (lack of preparation, short-term training, focus on obtaining practical results). As a rule, course participants do not have sufficient knowledge in the field of AI technologies. According to the "from simple to complex" learning method, the first laboratory work must be performed on specialized platforms with an intuitive interface (Orange Data Mining, Teachable Machine, etc.). Using such platforms makes it easier for untrained students to understand which tasks and algorithms are being solved within the framework of machine learning. The second method underlying the course program is a project–oriented approach. The project-based learning method, implemented by developing text-based Telegram bots or Telegram bots that identify images, allows students to gain practical knowledge useful in their professional activities. The AI literacy course program proposed in the article has proven its effectiveness in practice and contributed to the qualitative development of competencies in the field of artificial intelligence by students.

Keywords: artificial intelligence methods, AI literacy, widgets, text bots, python libraries, project learning, machine learning

Structural modeling of information interaction between producers and consumers in a digitalized organizational system

2025. T.13. № 4. id 2075

Pupykin A.N.

DOI: 10.26102/2310-6018/2025.51.4.042

This study investigates the optimization of management in a digitalized organizational system (DOS) through the lens of information interaction among participants in the ordering process. The primary focus is on the development of a structural model of the system and a management decision-making algorithm. The structural model of the organizational system is defined by the connections facilitated through an information interaction environment between the management center and the entities of producers, consumers, and aggregators. These connections are characterized by quantitative values of the DOS performance and resource indicators. One subset of these indicators is established as requirements, while the other is recorded during monitoring at specific time periods. The structural model of the management decision-making process is, firstly, defined by the combination of an expert mode and an intelligent support mode, and, secondly, by two stages: operation and development, within a given planning horizon. For these modes and stages, resource allocation management modules are substantiated: for allocating resources among consumer information flows during order formation; for allocating ergodic elements among non-ergodic elements of the DOS during order fulfillment; and for development conditions of the DOS within a given planning horizon. Intelligent support for management decisions is implemented through the integration of optimization, predictive modeling, and expert assessment. The foundation for managing the allocation of resources among information flows is a parametric optimization model, while the allocation of ergodic elements is based on a multi-alternative optimization model. For making decisions on the allocation of DOS development resources, the application of predictive models, trained via machine learning on retrospective monitoring data, is considered.

Keywords: digitalization, organizational system, structural modeling, management, optimization, forecasting

Fog Computing – the foundation for distributed digital project evolution and security in business

2025. T.13. № 4. id 2072

Titova I. Diakonov D.

DOI: 10.26102/2310-6018/2025.51.4.036

In the context of the growing complexity of communications, operational management, security systems and special attention is required to the infological tools and infrastructure of the organizations. The digital ecosystem is the basis of both communications, management impact, and the basis for the efficiency and safety of the quality of the company's systems and business processes. The purpose of the study is a system analysis of the modern computing infrastructure of distributed corporate work – vague or Fog Computing, its capabilities, technologies and problems, especially in the field of security. The methodology of the presented study includes methods of system analysis, information-logical modeling and expert heuristic approach. The results of the study were the identification of the range of problems of the foggy infrastructure, the possibilities of flexible and scalable increase in its capacity and security, fault tolerance, as well as approaches to solving systemic problems of an instrumental, informational, integration and other nature. A classification of foggy calculations was also carried out with an emphasis on their safety. Research can be further developed, in particular, in the direction of providing foggy computing infrastructures with the necessary degree of flexibility, standards and their integration, as well as building flexible models and algorithms for identifying process parameters.

Keywords: fog computing, security, distributed computing, digital design, enterprise infrastructure