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

Evaluating the potential of neural discrete choice model with siamese networks for a real estate sales forecasting problem

2025. T.13. № 2. id 1819

Razumovskiy L.G. Karenin N.E. Gerasimova M.A.

DOI: 10.26102/2310-6018/2025.49.2.021

The paper considers the problem of reproducing the process of purchasing real estate, the solution of which will allow testing both existing and future dynamic pricing algorithms, building predictions of buyers' preferences and forming a demand curve. As a solution, it is proposed to use an approach based on the use of discrete choice models, which are widely represented in the economic literature and have a wide range of applications in the field of studying consumer behavior and preferences in competitive markets. This paper presents a new discrete choice model that uses a neural network to form the utility of a real estate object. An approach to training the model through Siamese neural networks is proposed. The article also proposes a non-standard architecture of the main neural network, which allows avoiding the loss of convergence during its training. The paper simulates the process of purchasing real estate using classical models based on logistic regression with random coefficients and using a neural network model, and compares them. As a result of numerical experiments, a noticeable advantage of the proposed neural network approach is shown. Using a permutation test, the statistical significance of the obtained results is proved.

Keywords: discrete choice model, siamese neural networks, sales process, real estate, customer preference, econometric modeling

Mathematical model for constructing an idustry-specific career guidance decision support system

2025. T.13. № 1. id 1816

Stupina A.A. Osipov V.S. Bobyleva O.V. Yakovlev D.A.

DOI: 10.26102/2310-6018/2025.48.1.033

The article addresses the challenges of developing an industry-specific decision support system for education and career guidance in engineering professions under conditions of limited data availability. The system aims to facilitate informed career choices by assessing students’ aptitudes for engineering and technical fields. To formalize these aptitudes, the authors propose a set of key factors and evaluation metrics that enable data-driven conclusions using information extracted from digital educational environments. These factors are designed to leverage immersive technologies and digital educational tools for data acquisition. The study introduces a generalized mathematical model that quantifies the manifestation of multiple parameters and aligns them with potential professional trajectories. The model incorporates weighted indices and significance assessments for predictive analytics, along with methods to integrate diverse evaluation approaches into the decision support framework. Parameters include psychological diagnostics and academic performance metrics. Additionally, the paper demonstrates the application of the generalized model to the mining industry, validated through empirical testing involving a control group of industry professionals. The results highlight the model’s adaptability to sector-specific requirements and its capacity to enhance objectivity in career aptitude assessment. This research contributes to the development of scalable, data-informed tools for engineering career guidance, emphasizing the integration of emerging technologies into educational ecosystems.

Keywords: decision support systems, forecasting, mathematical modeling, data model, digital environment, career guidance

Complex of programs for determining the highest priorities of enrollees in the competitive lists

2025. T.13. № 1. id 1814

Baryshnikova N.Y. Vasin A.V. Galin A.V. Ratmanov A.S.

DOI: 10.26102/2310-6018/2025.48.1.025

The article discusses the issues of automation of the functionality of the admission campaign of the educational organization of higher education, in particular, issues related to the introduction of enrollment priorities. The enrollee applies for admission to higher education programs. In it, it denotes individual competitive groups and enrollment priorities for each of them. Based on the information provided, the educational organization of higher education determines the highest priorities for the further enrollment of enrollee. The complex of programs presented in this article is an urgent tool for solving the problem of automatically determining the highest priorities. The complex developed by the authors consists of two subprograms. Each subroutine contains its own implemented algorithm. One of the algorithms for solving the problem is an algorithm based on the use of the «brute force» method (the exhaustive search method). This method has proven its simplicity in implementation and readability of the code. Also, the Gale-Shapley algorithm is implemented in the complex of programs. It is characterized by the search for stable matchings between two groups of participants. Within the framework of this article, the main stages of the complex of programs are presented in detail. Finally, the authors analyzed the results of the implemented algorithms. It is concluded that the algorithms are effective. The results obtained in the article in the form of a complex of programs are proposed to be used by employees of admission commissions of educational institutions of higher education when conducting a new recruitment in terms of automation of determining the highest priorities of applicants in the competitive lists.

Keywords: complex of programs, «brute force» method, gale-Shapley algorithm, admission campaign, selection committee, enrollee, priorities, enrollment, stable matchings

Approximations of hospital statistics of recoveries from COVID-19

2025. T.13. № 1. id 1812

Borovsky A.V. Galkin A.L. Doroshenko S.S.

DOI: 10.26102/2310-6018/2025.48.1.024

Hospital statistics on COVID-19 recoveries in Irkutsk are presented in the form of the rate of recovery over a certain number of days from the full group of patients. The recovery time varies from 1 to 182 days. The number of cases considered reaches ~100000 cases. For the convenience of using the data, it is proposed to approximate the table for the recovery rate by various types of nonlinear functions. The following variants of approximating functions have been studied: Gaussian, Lorentz, modified Lorentz, Weibull function, Johnson functions. For comparison with statistics, methods were used to minimize the standard deviations of approximating functions from experimental data. The least squares method is used for functions with two and three parameters, the coordinate descent method, and the gradient descent method for functions with four fitting parameters. It is shown that the best fitting results are provided by a modified Lorentz function with four parameters. According to the degree of discrepancy with experimental statistics, the approximating functions are arranged in the following order: the Weibull function provides the least accurate fit (16.15%), followed by the Johnson function SU (10.65%), slightly better fit for the Johnson function SB (8.49%), for the Gaussian function (5.8%), for the Lorentz function the fit is (3.2828%), the best fit is given by the modified Lorentzian function (3.2804%) under certain approximations.

Keywords: epidemic theory, optimization methods, coordinate descent, gradient descent, least squares method, gauss approximation, lorentz approximation, weibull approximation, johnson approximation, modified Lorentz distribution

Detection of depression features with user data from social network using neural network

2025. T.13. № 1. id 1810

Solokhov T.D. Kochkarov A.A.

DOI: 10.26102/2310-6018/2025.48.1.020

The article studies the problem of identifying signs of depression based on user data from social networks using machine learning methods and network analysis. The study includes the development of a model for detecting users with signs of depression, which relies on text analysis of their social network posts and profile metadata. Neural networks were used as algorithms in the study, showing high classification accuracy. Network analysis was implemented to examine the influence of users with signs of depression and it shows that such users have low centrality and do not form dense clusters, indicating their social isolation. The hypothesis of depression spreading through social connections was not confirmed, suggesting minimal impact of depressive users on others. The research results can be utilized to develop systems for early detection of depression. Special attention is given to the study's limitations, including the use of data from a single social network and the complexity of processing textual data. The article proposes directions for further research aimed at expanding methods for analyzing the spread of depressive behavior in social networks.

Keywords: forecasting, depression, psychological disorder, classification, social network, machine learning, neural network, network analysis

Real-time monitoring of communication networks based on cloud computing

2025. T.13. № 1. id 1809

Amoa K. Sidorenko E.V. Ryndin N.A.

DOI: 10.26102/2310-6018/2025.48.1.014

When creating a communication network, various obstacles inevitably arise that negatively affect its effectiveness. The lack of measures to eliminate such interference makes it difficult to optimize the network. Among the problems caused by interference, the problem of blocking them is one of the most significant. This unresolved issue may make successful network design impossible. In order to solve the problems that the traditional method has a long response time to monitor the congestion of the communication network and the detection effect is not ideal, a real-time monitoring method based on cloud computing for blocking the communication network is proposed. Firstly, a communication network monitoring point is established, and the receiver completes the communication data collection process. Based on the collected data, continuous traffic calculation is performed to determine whether there is an emergency blocking state in the communication network channel and determine the exact location of the blocking point. In this way, the information generates an alarm message to obtain the monitoring results. The real-time running time and the accuracy of the monitoring method are experimentally analyzed. It is found that the monitoring method can control the delay time within 0.2 s, and the monitoring error rate is low.

Keywords: cloud computing, telecommunications, network congestion, real-time monitoring, monitoring point, system management, blocking

Neural network to optimize the adaptive exponential min sum decoding algorithm

2025. T.13. № 1. id 1807

Zhang W. Mouhamad I. Saklakov V.M. Jayakody D.K.

DOI: 10.26102/2310-6018/2025.48.1.026

Currently, deep learning, as a hot research direction, has yielded fruitful research results in natural language processing and graph recognition and generation, such as ChatGPT and Sora. Combining deep learning with decoding algorithms for channel coding has also gradually become a research hotspot in the field of communication. In this paper, we use deep learning to improve the adaptive exponential min sum (AEMS) algorithm for LDPC codes. Initially, we extend the iterative decoding procedure between check nodes (CNs) and variable nodes (VNs) in the AEMS decoding algorithm into a feedforward propagation network based on the Tanner graph derived from the H matrix of LDPC codes. Second, in order to improve the model training efficiency and reduce the computational complexity, we assign the same weight factor to all the edge information in each iteration of the AEMS decoding network, which reduces the computational complexity while guaranteeing the decoding performance, and we call it the shared neural AEMS (SNAEMS) decoding network. The simulation results show that the decoding performance of the proposed SNAEMS decoding network outperforms that of the conventional AEMS decoder, and its coding gain is gradually enhanced as the code length increases.

Keywords: LDPC, deep learning, neural network, exponential algorithm, min sum

Sensor data integration system in onboard control systems of unmanned aerial systems

2025. T.13. № 1. id 1806

Guliutin N.N. Ermienko N.A. Antamoshkin O.A.

DOI: 10.26102/2310-6018/2025.48.1.019

Modern unmanned aerial systems (UAS) play a key role in various industries, including environmental monitoring, geodesy, agriculture, and forestry. One of the most critical factors for their successful application is the integration of data from various sensors, such as global navigation satellite systems, inertial navigation systems, lidars, cameras, and thermal imagers. Sensor data fusion significantly enhances the accuracy, reliability, and functionality of control systems. This paper explores data integration methods, including traditional algorithms like Kalman filters and their extended versions, as well as modern approaches based on deep learning models, such as FusionNet and Deep Sensor Fusion. Experimental studies have shown that learning-based models outperform traditional algorithms, achieving up to a 40 % improvement in navigation accuracy and enhanced resilience to noise and external disturbances. The proposed approaches demonstrate the potential to expand UAS applications in autonomous navigation, cartography, and monitoring, particularly in challenging operational environments. Future development prospects include the implementation of hyperspectral sensors and the development of adaptive data integration methods to further improve the efficiency and effectiveness of unmanned systems.

Keywords: sensor data integration, unmanned aerial systems, kalman filter, fusionNet, deep Sensor Fusion, autonomous navigation, resilience to disturbances

On achievability of consensus in multi-agent control systems with a leader

2025. T.13. № 1. id 1805

Yang S.

DOI: 10.26102/2310-6018/2025.48.1.023

The paper proposes a distributed control algorithm for multi-agent systems with a leader. The main objective is to ensure the asymptotic convergence of the states of all follower agents to the state of the leader, under the condition that each agent uses only local information obtained from neighboring nodes. The dynamics of the agents are modeled by a second-order system – a double integrator, which allows to take into account both the position and velocity of the agents. This description more accurately reflects the properties of real systems compared to the commonly used simplified first-order models. Graph theory is employed to formalize the topology of communication links between agents. The developed algorithm is based on the idea of pinning control and uses local information about the states of neighboring agents and the leader. The Lyapunov method and eigenvalue analysis were used to study the stability of the system and to obtain analytical conditions for the gain factors that guarantee the achievement of consensus. To illustrate the efficiency and effectiveness of the proposed algorithm, numerical simulations are conducted in MATLAB. The leader's trajectory is chosen based on the optimal trajectory obtained in previous studies by the authors. The results confirm that the states of the follower agents asymptotically converge to the state of the leader over time. The proposed algorithm can be applied to solve problems of group control of mobile robots, unmanned vehicles, and other distributed technical systems.

Keywords: multi-agent systems, distributed control, consensus, leader-follower structure, graph theory, pinning control, group control

Development of API rate limiting methods based on consumer classes

2025. T.13. № 1. id 1803

Seleznev R.M.

DOI: 10.26102/2310-6018/2025.48.1.013

Rate limiting is a crucial aspect of managing the availability and reliability of APIs. Today, there are several approaches to implementing rate limiting mechanisms, each based on specific algorithms or their combinations. However, existing methods often treat all consumers as a homogeneous group, hindering the creation of flexible resource management strategies in modern distributed architectures. In this article, the author proposes two new methods for rate limiting based on the token bucket algorithm. The first method involves using a shared token bucket with different minimum fill requirements depending on the consumer class. The second method suggests using separate token buckets for each consumer class with individual parameter values but a common limit. Simulation results confirmed that both methods enable efficient API request limitation, though disparities emerged regarding resource distribution patterns across diverse consumer classes. These findings have practical implications for developers of information systems and services who need to maintain high availability while ensuring access guarantees for various consumer categories.

Keywords: rate limiting, token bucket algorithm, software interface, consumer class, quota, threshold, burst traffic

A method for generating closed-type questions using LLMs

2025. T.13. № 1. id 1799

Dagaev A.E.

DOI: 10.26102/2310-6018/2025.48.1.021

This study presents a method for closed-ended question generation leveraging large language models (LLM) to improve the quality and relevance of generated questions. The proposed framework combines the stages of generation, verification, and refinement, which allows for the improvement of low-quality questions through feedback rather than simply discarding them. The method was tested on three widely recognized datasets: SQuAD, Natural Questions, and RACE. Key evaluation metrics, including ROUGE, BLEU, and METEOR, consistently showed performance gains across all tested models. Four LLM configurations were used: O1, O1-mini, GPT-4o, and GPT-4o-mini, with O1 achieving the highest results across all datasets and metrics. Expert evaluation revealed an accuracy improvement of up to 14.4% compared to generation without verification and refinement. The results highlight the method's effectiveness in ensuring greater clarity, factual correctness, and contextual relevance in generated questions. The combination of automated verification and refinement further enhances outcomes, showcasing the potential of LLMs to refine text generation tasks. These findings will benefit researchers in natural language processing, educational technology, and professionals working on adaptive learning systems and corporate training software.

Keywords: question generation, large language models, artificial intelligence, natural language processing, o1, o1-mini, GPT-4o, GPT-4o-mini

Automated user segmentation using RFM analysis in marketing strategies

2025. T.13. № 1. id 1798

Svyatov R.S.

DOI: 10.26102/2310-6018/2025.48.1.018

The relevance of the study is determined by the need to enhance the effectiveness of marketing strategies through automated and customizable customer segmentation. This work proposes a universal customer data management system based on RFM segmentation with the ability to configure flexible logic, as well as the capability to integrate with various external systems. Traditional CRM systems and manual RFM segmentation methods are limited in functionality and do not always meet the business needs for flexibility and integration with various data sources. The study identifies the shortcomings of traditional CRM systems and suggests points for improvement in the described system. Additionally, an experiment was conducted comparing the RFM segments generated using the proposed architecture with Yandex's auto-strategies in the Yandex.Direct advertising platform. The application of the system showed significant advantages over auto-strategies, including a 30.71% increase in purchases in the case of a clothing store. The results confirm the practical value of the system for optimizing marketing campaigns and improving conversion. The results are of practical importance for companies in need of customized solutions and integrations. Further development is proposed, focusing on improving the RFM segmentation method by implementing machine learning algorithms and exploring additional effective channels for utilizing the generated segments.

Keywords: RFM analysis, marketing automation, customer loyalty, user segmentation, e-commerce, advertising strategy optimization

System analysis and modeling of the profitability of the energy service contract based on the digital ruble

2025. T.13. № 1. id 1797

Kaziev V.M. Kazieva B.V.

DOI: 10.26102/2310-6018/2025.48.1.015

Enterprises participating in housing and communal services need market energy viability and competitiveness, attractiveness for consumers. For Russian companies, it is important to adhere to relatively "soft" (flexible) tariffs and energy supply strategies. It is necessary to find effective solutions, for example, investment and reducing uncertainties such as "white noise" in the energy system. The purpose of the study is a systematic analysis of the potential of a smart contract, a digital ruble and digital payments in energy service contracts. The possibilities of energy contracts and services, as well as the content and features of such contracts, measures for sustainable energy conservation with a certain profitability and optimization of energy resources were studied by methods of system analysis and modeling. Therefore, it is necessary to identify the parameters and features of the contract and simulate the processes of energy supply. The results of the study are: 1) a systematic analysis of standard forms of contracts and a description of a set of energy-saving key procedures of the enterprise; 2) analysis of the potential of the digital ruble and its "energy capabilities"; 3) model of dynamics of management of an energy service enterprise based on diffusion of digital services and its research. The results of the work will expand the possibilities of concluding and developing energy service contracts in practice, as well as build flexible models and algorithms for energy supply.

Keywords: system analysis, smart contract, energy consumption, energy service contract, modeling

A system for analyzing images of nucleated bone marrow cells for the formation of a diagnostic conclusion in oncohematology

2025. T.13. № 1. id 1796

Polyakov E.V. Popov V.V. Dmitrieva V.V.

DOI: 10.26102/2310-6018/2025.48.1.016

The paper presents a system for analyzing images of nucleated bone marrow cells to form a diagnostic conclusion in oncohematology, aimed at solving the problem of constructing a data processing pipeline in automatic analyzers of biomedical images. The relevance of the study is due to the need to improve the reliability of theof automatic microscopic analysis of biomedical samples, which is aa difficult task due to high variability and morphological complexity of the investigated objects. One solution to this problem is to develop a web service that uploads, processes and describes images, then classifies them into categories of confirmed and unconfirmed cases. This web service provides cross-platform and accessibility, builds an open database of verified images and providestools for processing and analyzing images, as well as tools for correcting by the physician of the processing results. The system does not prescribe treatment and does not make diagnoses in dependently, but serves as an intelligent tool for processing, analyzing and transmission of research results in real time. The testing results showed high accuracy of the system: 91% for neural network methods and up to 97% for classical algorithms. The developed system allows for the analysis of data processing modules for computer microscopy systems.

Keywords: analysis of biomedical images, selection of objects, classification of nucleated cells, pattern recognition, oncohematology

Application of elitist ant system and Max-Min ant system algorithms for path optimization in quantum key distribution networks

2025. T.13. № 1. id 1792

Razdyakonov E.S. Korchagin S.A. Timoshenko A.V. Bulatov M.F.

DOI: 10.26102/2310-6018/2025.48.1.012

This study focuses on route optimization in quantum key distribution (QKD) networks, whose features are a number of physical constraints and strong topology dependence. This paper examines the application of two variations of the ant colony algorithm, the elitist ant system (EAS) and Max-Min ant system (MMAS) algorithms, to construct optimal routes in QKD networks. A metric for the communication efficiency of a route in QKD networks has been presented to evaluate the quality of a route according to given capacity and security requirements. The peculiarity of this metric is its non-additive capacity component, which depends on the minimum link efficiency in the route. A series of experiments were conducted on a randomly generated planar graph for long and short routes with EAS and MMAS algorithms, which resulted in MMAS being significantly more efficient for long routes, but in the case of short routes, EAS found the route faster without significant loss in solution quality. The results obtained in this study can be applied in solving problems of dynamic routing, as well as optimization of the topology of quantum key distribution networks.

Keywords: quantum key distribution, metaheuristics, ant algorithm, elitist ant system, max-Min ant system, pathfinding

Simplified physical models of the mediastinum of newborns for electrical impedance tomography

2025. T.13. № 1. id 1791

Konko M.A. Aleksanyan G.K. Gorbatenko N.I. Elkin N.S. Temnyakov N.S.

DOI: 10.26102/2310-6018/2025.48.1.011

The article presents the results of the development and experimental study of two simplified physical models of the neonatal mediastinum for electrical impedance tomography. The created models are based on spiral computed tomography data and take into account the anatomical features of the infant chest organs. The designs were implemented using 3D printing technologies, which made it possible to achieve high accuracy of geometric parameters. The models are equipped with a controlled air filling system for the lungs and three rows of electrodes, which makes it possible to conduct experiments on modeling global and regional ventilation. Experimental studies have demonstrated that the developed models make it possible to record respiratory volumes in the range from 2 to 120 ml, which corresponds to the physiological parameters of newborn breathing. The data obtained confirmed the operability of the models, their sensitivity to changes in air volumes, as well as their suitability for research and testing of new algorithms and methods in the field of electrical impedance tomography. It was found that the proposed models provide adequate reproduction of ventilation processes and can be used to develop diagnostic solutions in the field of neonatology. The results of the work are of practical value for scientific research aimed at improving methods for diagnosing respiratory disorders in newborns, and can be used in educational practice.

Keywords: simplified physical model of mediastinum, electrical impedance tomography, newborns, process of global and regional ventilation, lungs

Determination of the optimal geometric parameters of the metal seal of the tubing hanger

2025. T.13. № 1. id 1789

Timofeev E. Godenko A. Tarasova I.

DOI: 10.26102/2310-6018/2025.48.1.017

Operation of underwater complex hydrocarbon production systems is accompanied by increased risks of emergency situations, in particular, leaks and emissions due to loss of sealing between the underwater fountain fittings and the tubing hanger. Emissions and leaks during operation of underwater wells can lead to such irreversible consequences as loss of produced products and harm to the environment, as well as damage to expensive equipment, which requires expensive and technically complex repairs. For this reason, in the process of experimental design work on the design of this type of equipment, it is necessary to carry out high-quality and effective calculation support for the developed metal seals, allowing for determining their optimal geometry. The authors have developed a mathematical model for determining the stress-strain state of a metal seal of the tubing hanger, taking into account its rigidity characteristics. To determine the optimal geometry of a metal seal, the theory of qualities was used, based on which the overall quality of a metal seal was assessed by its strength and tightness. For the tightness and strength of a metal seal, particular quality functions were constructed, which are combined into an objective function using the Kolmagorov-Nagumo functional average. The results of optimizing the standard geometric parameters of a metal seal of a prototype of a tubing hanger according to the proposed method are presented. The information contained in this publication is useful to engineering and scientific professionals involved in the development and research of methods for ensuring the tightness of underwater connections using metallic seals.

Keywords: subsea production system, metal seal, stress-strain state, tubing hanger, underwater fountain fittings, contact pressure, quality theory, optimization

Biotechnical system of personalized rehabilitation of patients with limited motor functions

2025. T.13. № 1. id 1787

Filist S.A. Petrunina E.V. Pshenichny A.E. Ermakov D.A. Krupchatnikov R.A. Serebrovskiy V.V.

DOI: 10.26102/2310-6018/2025.48.1.002

The article considers a rehabilitation biotechnical system with an adaptable virtual reality intended for rehabilitation of patients with impaired motor functions of the lower limbs in rehabilitation complexes with combined feedback. The biotechnical system has the following functional modules: formation of controlled effects on the patient, control of controlled effects, rehabilitation management and information support. During rehabilitation, the patient's muscle fatigue and its dynamics are monitored. This made it possible to make adjustments to the rehabilitation block program during a rehabilitation session and manage the procedure for adapting virtual reality to the patient's functional state, as well as to carry out mathematical modeling of rehabilitation course scenarios. A model for planning a rehabilitation course using biofeedback intended for a biotechnical system with virtual reality is proposed. An experimental group was formed to assess the effectiveness of rehabilitation of post-stroke patients with paretic lower limbs. The rehabilitation results in this group showed that the choice of virtual reality content adapted to the patient allows increasing the effectiveness of rehabilitation according to the LEFS scale by 11%. Experimental studies of the effectiveness of muscle fatigue control during rehabilitation have been conducted. It is confirmed by testing the statolocomotor sphere according to the Tinetti scale, the indicators of which, on average, exceeded the indicators in the comparison group by 10%. Inclusion of adaptive virtual reality and muscle fatigue monitoring in the rehabilitation process leads to earlier restoration of impaired balance function, motor activity and social rehabilitation.

Keywords: persons with limited mobility, biotechnical system, impaired mobility, virtual reality, biofeedback, muscle fatigue

Intelligent system for managing physiological load based on IoT devices and data processing methods

2025. T.13. № 1. id 1786

Mikhalev A.S. Podolyak A. Golovenkin S.E. Savitsky I.V. Antamoshkin O.A.

DOI: 10.26102/2310-6018/2025.48.1.029

With the growing popularity of wearable IoT devices for cardiovascular monitoring, their use faces the problem of measurement accuracy, especially during physical activity. This paper focuses on developing a methodology for detecting and eliminating anomalies in heart rate (HR) data collected from IoT devices to assess myocardial workload. As part of the work, an experiment was conducted in which HR data collected from wearable IoT devices (smart watches) were compared with the readings of certified medical equipment (Holter monitor). An algorithm for time series analysis is proposed, including the stages of data preprocessing, anomaly detection and correction them. Isolation forest algorithm was used to detect anomalies. The results of the study demonstrated that the proposed approach can reduce measurement error and achieve acceptable accuracy in the range of HR 90–120 beats per minute, which is critical for cardiac rehabilitation tasks. Based on the cleaned data, a model for classifying physical activity levels was developed, including recommendations for optimizing the patient's activity. The proposed methodology combines elements of system analysis, control and information processing, which makes it universal for application in intelligent health monitoring systems. The obtained results emphasize the prospects of IoT devices as a basis for building remote cardiac rehabilitation systems that can improve the quality of life of patients and reduce the burden on healthcare.

Keywords: ioT devices, heart rate, anomaly detection, intelligent data analysis, load management, cardiorehabilitation

Structural modeling for management in organizational systems with a heterogeneous structure of spatial elements

2024. T.12. № 4. id 1785

Linkina A.V.

DOI: 10.26102/2310-6018/2024.47.4.041

The article discusses the process of structural modeling in management in organizational systems with a heterogeneous structure of spatial elements. It is shown that the objects under study belong to territorially distributed organizational systems. Their peculiarity is the variety of spatial elements within a limited territory, which significantly influence the indicators of the effective functioning of the activity environment. Three classes of structural models are identified: object, system and process levels. The processes of structural modeling of the organizational system, its management, and management decision-making are examined in detail. It has been determined that the structural model of the organizational system reflects the interaction of the control center, objects with a heterogeneous structure of spatial elements, and a geographically distributed environment that unites objects into an organizational whole. The structural model of the management system is focused on the balanced interaction of the traditional management subsystem based on expert opinions and the decision support subsystem based on optimization of resource distribution, transformation and variation processes. The structural model of the process level is formed as an invariant sequence of algorithmic actions when making management decisions. It is substantiated that in order to meet the requirements of the control center it is necessary to combine algorithmic actions within the following stages: identification of classification signs of heterogeneity in the structure of spatial elements of objects, expert assessment of quantitative parameters for the formation of a multivariate choice model, generation of options for management decisions, expert selection on the generated set.

Keywords: organizational system, management, heterogeneity of the structure of spatial elements, structural modeling, expert optimization modeling

Optimization of corporate learning process management: a mathematical model and its application

2025. T.13. № 1. id 1782

Kharitonov I.A.

DOI: 10.26102/2310-6018/2025.48.1.006

This article considers the problem of optimization of the employee training process control at an enterprise, including the distribution of teachers, students, lessons and training rooms under multiple constraints. The relevance of the work is determined by the need for effective management of training processes in organizations, considering the qualifications of teachers, skills of employees, time constraints and the sequence of skill acquisition. To formalize the problem, a mathematical model was developed that allows for a linear description of the key aspects of training. The model includes multidimensional constraints such as training time, teacher employment, availability of training places and the sequence of skill acquisition. However, due to the combinatorial nature of the problem and discrete variables, its solution requires the use of specialized methods. To solve the problem, optimization approaches are used that include: formalization of the problem in a linear manner to identify subtasks that can be solved separately (for example, determining available classes for teachers); application of heuristic methods and dynamic programming for the final distribution of classes and resources. The proposed model demonstrates its effectiveness in personnel training management scenarios where both cost minimization and fulfillment of all specified constraints are important. Despite the limitations of the linear description, the model provides a simplification of the solution due to the structured approach to resource allocation. This makes it a universal tool applicable to the management of employee training in companies of various types. The materials of the article can be useful for the development of adaptive learning management systems, as well as for further research aimed at improving resource allocation algorithms.

Keywords: in-house training, mentoring, training optimization, resource planning, skills sequencing, human resources management, employee qualifications, training simulation

Mathematical models and software complex for intelligent analysis and forecasting the performance of government contracts

2025. T.13. № 1. id 1778

Rubtsov D.Y.

DOI: 10.26102/2310-6018/2025.48.1.010

The paper proposes mathematical models and a software package for intellectual analysis and forecasting of the execution of government contracts, based on a neural network and classical machine learning methods trained on a retrospective database of counterparties and contracts. A set of mathematical models and programs allows you to calculate the probabilities and risks of non-fulfillment of government contracts, thereby reducing budget losses and positively influencing the stability of the real sector of the economy. A comparative analysis of machine learning methods was carried out: logistic regression, decision tree, support vector machine and neural network model. A model has been developed that allows forecasting with an accuracy of 97.89%. For each mathematical model, a separate module has been developed, which together constitute a software package. The neural network model showed a result of 87.65%, which is associated with a relatively small set of data for training; however, this model allows us to reveal the further potential of the system in connection with continuous training in real time on new contracts, for the evaluation of which the proposed software package will be used. The results of the study can be used to further improve decision support systems in the field of procurement and its application in order to improve the overall quality of analysis and forecasting of the implementation of government contracts.

Keywords: mathematical modeling, software package, data analysis, government contracts, machine learning, intelligent system, forecasting

Dynamic model of coordinateometry of nearby signal sources using the super-resolution algorithm MUSIC

2025. T.13. № 1. id 1775

Glushankov E.I. Kondrshov Z.K. Syrovetnik D.S. Rylov E.A.

DOI: 10.26102/2310-6018/2025.48.1.004

In recent years, the problem of radar and radio direction finding has become very relevant due to the rapid development of microelectronics for small-sized unmanned aerial vehicles and communication terminals. The article is devoted to determining the coordinates of signal sources. In particular, closely spaced, uncorrelated emitters and receiving antenna arrays on several mobile devices spaced apart in space are investigated. To resolve such signal sources, the MUSIC super-resolution algorithm is used, followed by solving the coordinate measurement problem using the least squares method. A software model was created in the MATLAB environment, implementing a dynamic system in which each radio device has its own trajectories and speed indicators. A comparative analysis of the accuracy of the results obtained in various situations from the point of view of geometry and dynamics was carried out. It was found that the algorithm works most effectively in the case of targets located inside the zone formed by the scanning objects. In this case, the accuracy of determining coordinates is comparable to the distance between the signal sources. Based on the obtained results, it is possible to construct radar receivers for direction finding of nearby signal sources in the main lobe of the directional pattern of the receiving antenna of a location station, which is required when solving a number of location and monitoring problems in a complex electronic environment.

Keywords: direction finding, coordinateometry, super-resolution, MUSIC, MATLAB

A comprehensive analysis of the impact of G1 Garbage Collector parameters on JVM performance and stability

2024. T.12. № 4. id 1774

Zolotukhina D.Y.

DOI: 10.26102/2310-6018/2024.47.4.039

The relevance of thе study is determined by the necessity of improving memory management efficiency in high-load Java applications, where minimizing garbage collection pauses and maintaining high throughput are critically important tasks. This article aims to systematically investigate the parameter space of the G1 Garbage Collector (G1 GC) and develop practical recommendations for its optimization under high-load conditions. The primary research method is an empirical approach, which involves the development of a multithreaded Java application capable of generating sustained memory and CPU loads. The study utilized a control and six experimental G1 GC configurations, differing in parameters such as heap region size, heap occupancy threshold, maximum pause duration, young region size, the number of GC threads, and the activation of Periodic GC. Key metrics, including pause durations, GC frequency, throughput, and freed memory volume, were measured, visualized, and systematically analyzed using the GCViewer tool. The article presents recommendations for G1 GC optimization, highlights the advantages of reallocating memory toward young regions and enabling Periodic GC, and identifies the limitations of the MaxGCPauseMillis parameter in aggressive configurations. The findings have practical value for developers of high-load applications requiring low latency and high system stability. The conclusions contribute to a deeper understanding of G1 GC functionality and can serve as a foundation for further research in JVM memory management.

Keywords: g1 garbage collector, memory management, jvm optimization, high-load applications, gc parameter tuning, throughput, pause duration, garbage collection, young memory regions, java application performance

The role of machine learning algorithms in optimizing agricultural production: a review of international experience and adaptation to Ethiopian conditions

2025. T.13. № 1. id 1773

Mekecha B. Gorbatov A.V.

DOI: 10.26102/2310-6018/2025.48.1.005

The relevance of this study is driven by the need to enhance the efficiency of agricultural production in response to the growing demand for food security, particularly in economically underdeveloped countries such as Ethiopia. The main objective of the research is to explore the potential application of machine learning algorithms to optimize agricultural processes and adapt international practices to the specific conditions of Ethiopia. The methodological approach includes an analysis of contemporary scientific literature on the use of machine learning in agriculture and the systematization of successful practices involving algorithms such as CNN, LSTM, RNN, and Q-Learning. The study investigates the characteristics of Ethiopia's agricultural sector, including existing barriers to the adoption of advanced technologies. The results highlight that machine learning algorithms hold significant potential for increasing crop yields, improving soil and crop monitoring, and forecasting climate risks. Specifically, utilizing data from drones and sensors enables the creation of precise models for managing agricultural processes. However, key challenges such as insufficient funding, a lack of specialized data processing infrastructure, and limited access to technology have been identified. The study concludes by emphasizing the importance of attracting governmental and international investments, developing tailored databases, and creating models that account for local conditions. The findings provide practical value for developing strategies to digitize agriculture and prevent food crises in countries facing similar challenges.

Keywords: machine learning, artificial intelligence, agricultural production, precision agriculture technologies, ethiopian conditions

Method of intelligent decision support for forming tutor’s assistant group for checking free response works in online learning management

2024. T.12. № 4. id 1769

Latypova V.A.

DOI: 10.26102/2310-6018/2024.47.4.031

Despite the attractiveness of massive online courses for learners, only a small part of the latter reaches the finish line. This situation arises due to the effects of different adverse factors on the learning process. Additional staff and “smart” assistants (educational chatbots) are used for reducing the impact of these factors by helping tutors in online learning management. Tutor’s assistants are engaged to perform checking free response works and identifying course problems connected with its content, and educational chatbots to guide students through a course and organize interactions between them. On every launch of the online course, a tutor faces a choice of a group of the most suitable assistants. In the existing research, different assistants’ features such as: scores, motivation, way of communicating, etc. are taken into consideration in this selection. However, in the former, the ability of assistants to properly assess and comment on the performance of free response works are not taken into account. To close the gap, a method of intelligent decision support for forming tutor’s assistant group for checking such works is proposed in the paper. The method was tested on one of the laboratory works on the course “Modeling” and allowed to form a group of assistants capable of assessing the work correctly.

Keywords: intelligent decision support, tutor’s assistant, feedback mining, online learning management, free response work, online course

About a decentralized solution to the problem of simultaneous arrival of autonomous mobile objects to the final point using large data analysis

2024. T.12. № 4. id 1768

Chernoivanenko I.A. Kravets O.Y.

DOI: 10.26102/2310-6018/2024.43.4.036

The need to switch to more advanced control methods when using a conventional autonomous mobile facility (AMO) to control simultaneous arrivals arises due to excessive deviation. An innovative solution to this problem is the use of a decentralized management method to control the simultaneous arrival of the AMO to the final point, which is based on the analysis of big data. A solution was proposed to combine decentralized information through the use of filtering, on the basis of which the decentralized coordination of formations is managed. The article presents the main characteristics of AMO, shows the parameters of combining information about AMO, describes decentralized coordination management of formation and calculates the optimal path and convergence rate for decentralized management, and also takes into account restrictions on communication delay. An experimental study of errors in the x direction by the proposed method was carried out and compared with errors in the experiment without using this control method. Graphs comparing the convergence rate are also presented. The results of the experiment showed that the decentralized management method has a significant impact on the definition of the aim of AMO and the convergence of errors. Thanks to the proposed approach, it was possible to increase the efficiency of management and reduce errors, thereby proving the expediency of using this management method.

Keywords: large data analysis, autonomous mobile objects, decentralized management, information filtering, coordination management of formation

The role of user identification in predicting target actions on a website

2024. T.12. № 4. id 1767

Svyatov R.S.

DOI: 10.26102/2310-6018/2024.47.4.037

The relevance of study lies in the need to improve the accuracy of predicting users' target actions on websites, which is a key aspect of optimizing marketing strategies and personalizing user experiences. The complexity of the task is exacerbated by the lack of stable identifiers, leading to data fragmentation and reduced prediction accuracy. This paper aims to analyze the impact of user identification methods and develop approaches to segmentation, which will help to eliminate existing gaps in this area. The primary research method involves applying machine learning algorithms to evaluate the influence of different identifiers, such as client_id and user_id, on prediction accuracy. Segmentation of users was carried out based on the gradient boosting method, as well as an analysis of the effectiveness of retargeting campaigns in the Yandex.Direct system based on conversion rates, customer acquisition costs, and the share of advertising expenses using the example of a client specializing in the sale of e-books. The findings reveal that utilizing the user_id identifier improves purchase prediction accuracy by 8%, recall by 6%, and the F1-score by 7%. Segmenting users into targeted groups demonstrated a 67% reduction in customer acquisition cost, a decrease in advertising expense share to 5.87% compared to Yandex auto-strategies, and an increase in conversion rate to 34%. The article's materials are of significance for specialists in the field of e-commerce and marketing, providing a scientific basis for the implementation of personalized advertising campaigns. The proposed methods also offer potential for further enhancement of analytics and data integration in multichannel environments.

Keywords: machine learning, user behavior analysis, user identification, user segmentation, e-commerce, target action prediction

Systematization of CT reconstruction filters for artificial intelligence technologies: a retrospective study for chest and brain

2025. T.13. № 1. id 1766

Vasilev Y.A. Blokhin I.A. Gonchar A.P. Kodenko M.R. Reshetnikov R.V. Arzamasov K.M. Omelanskaya O.V.

DOI: 10.26102/2310-6018/2025.48.1.003

The selected convolution kernel in computed tomography (CT) directly affects the results of artificial intelligence (AI) algorithms. The formation of uniform requirements for this parameter is complicated by the fact that such filters are unique to equipment developers. The aim of the work is to create a table of correspondence of reconstruction filters between different equipment manufacturers to direct to the AI algorithms the series of images on which, in CT of the chest organs and the brain, the quantitative analysis will be most reproducible. DICOM tags 0018,1210 (Convolution Kernel), 0008,0070 (Manufacturer), 0018,0050 (Slice Thickness) of CT images from the Unified Radiology Information Service of Moscow were downloaded and analyzed. Inclusion criteria: age older than 18 years; slice thickness ≤ 3 mm. The data analysis is presented in the form of summary tables comparing reconstruction filters from different manufacturers for chest and brain CT, a number of clinical tasks, as well as descriptive statistics of their distribution by scanning area and manufacturer. 1905 chest ("CHEST" and "LUNG") and brain ("HEAD", "BRAIN") CT studies were included in the analysis. In chest CT, reconstructions to evaluate pulmonary parenchyma and mediastinal structures were common. Reconstructions for brain parenchyma and bone structures were common in brain CT. Systematization of reconstruction filters for chest and brain CT was performed. The obtained data will allow correct image series selection for quantitative AI analysis.

Keywords: reconstruction filters, computed tomography, artificial intelligence, chest organs, brain, data systematization

Analysis of the stability of water-saturated soils under cyclic influence: mathematical models and forecasts

2025. T.13. № 1. id 1765

Tishin N.R. Ozmidov O.R. Proletarsky A.V.

DOI: 10.26102/2310-6018/2025.48.1.001

The article discusses the mathematical modeling of soil liquefaction under the influence of dynamic loads, such as seismic, storm, or technogenic cyclic impacts. The liquefaction process, in which soil loses strength and bearing capacity, is critical for assessing the safety of construction objects, especially in areas with increased seismic activity or water-saturated soils. Several approaches were used for modeling, including the following functions: the exponential function from the work of H. Bilge et al. (2009), the logarithmic function from the work of V. Lentini et al. (2018), the power function (polynomial) proposed by C. Guoxing et al. (2018), an additional logarithmic function from the study of E. Meziane et al. (2021), and a hyperbolic function proposed by the authors of this article, which approximated the soil's resistance to cyclic impacts. The study analyzed laboratory test data for various soil types, combined into engineering-geological elements. Each function was analyzed in terms of approximation accuracy using the least squares method, which minimized the deviations between experimental and theoretical values. When evaluating the functions, consideration was given to how each behaves under a large number of loading cycles, which is important for predicting liquefaction under intense and prolonged loads. The selection of the optimal function was made by comparing the MSE and R2 metrics presented in the results tables. The application of the research results has practical significance in geotechnical design, especially for calculating foundations and underground structures in conditions of potentially liquefiable soils. Choosing the most suitable function for modeling soil liquefaction allows predicting soil stability under long-term and intense cyclic loads, minimizing the risk of deformation and destruction of structures.

Keywords: soil liquefaction, mathematical modeling, geotechnical engineering, dynamic loads, soil liquefaction function, liquefaction potential