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

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

Evaluation of the quality of intelligent text paraphrasing in Russian

2024. T.12. № 4. id 1763

Dagaev A.E. Popov D.I.

DOI: 10.26102/2310-6018/2024.47.4.038

The study focuses on the development of an integral metric for evaluating the quality of text paraphrasing models, addressing the pressing need for comprehensive and objective evaluation methods. Unlike previous research, which predominantly focuses on English-language datasets, this study emphasizes Russian-language datasets, which have remained underexplored until now. The inclusion of datasets such as Gazeta, XL-Sum, and WikiLingua (for Russian) as well as CNN Dailymail and XSum (for English) ensures the multilingual applicability of the proposed approach. The proposed metric combines lexical (ROUGE, BLEU), structural (ROUGE-L), and semantic (BERTScore, METEOR, BLEURT) evaluation criteria, with weights assigned based on the importance of each metric. The results highlight the superiority of ChatGPT-4 on Russian datasets and GigaChat on English datasets, whereas models such as Gemini and YouChat exhibit limited capabilities in achieving semantic accuracy regardless of the dataset language. The originality of this research lies in the integration of multiple metrics into a unified system, enabling more objective and comprehensive comparisons of language models. The study contributes to the field of natural language processing by providing a tool for assessing the quality of language models.

Keywords: natural language processing, text paraphrasing, gigaChat, yandexGPT 2, chatGPT-3.5, chatGPT-4, gemini, bing AI, youChat, mistral Large

Dynamic pricing for real estate

2025. T.13. № 1. id 1761

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

DOI: 10.26102/2310-6018/2025.48.1.008

The paper considers a mathematical model for dynamic price adjustment in real estate. The model is characterized by a finite number of real estate objects, a fixed sales horizon, and the presence of intermediate goals for sales and revenue. The model developed in this work addresses the case of variable total demand, incorporating the time value of money and the increase in real estate objects value as construction progresses. The general structure of pricing policy is studied, and an algorithm for determining prices under variable total demand is presented. Similar constructions are carried out for a model that accounts for the time value of money and the rising property value during construction. The case of a linear elasticity function is also examined as a basic but widely used practical scenario. Rigorous mathematical proofs of the results are provided, along with numerical simulations based on real estate data from a specific city over 3.5 years to compare different approaches to pricing policy formulation. The obtained results can be applied to effectively manage real estate pricing.

Keywords: dynamic pricing, real estate, price adjustment, variable total demand, cost of money, price increase, stages of construction

Multi criteria decision-making with the use of ELECTRE methods group

2024. T.12. № 4. id 1760

Latypova V.A.

DOI: 10.26102/2310-6018/2024.47.4.030

Solving semi-structured problems is an essential part of the organizational system management. To simplify addressing these problems, different methods of multi-criteria decision-making are used. Among the basic widespread methods can be distinguished ELECTRE family methods. A large number of scientific works are devoted to the latter, but nevertheless they do not give enough coverage to the following problem: when using different ELECTRE methods for solving the same task, you can get an unequal result. The reason is that these methods possess their own specific features along with the common basis. A method of multi criteria decision-making with the use of a group of ELECTRE methods: ELECTRE I, ELECTRE Iv, ELECTRE Is, ELECTRE II, ELECTRE III, ELECTRE IV; considering results of each method and applying integral scores of alternatives in defining an overall task solution, is suggested in the paper to eliminate the problem. The proposed method has been validated on a test case of multi-criteria selection of a candidate for a vacant position in the hiring process in human resource management. The former allowed to smooth out the discrepancies in the results each of the methods of the group and identify a comprehensive solution.

Keywords: decision-making, ELECTRE, multicriteria choice, integral score, expert assessment

Mathematical formalization of agent conflict in achieving local goals

2024. T.12. № 4. id 1757

Rossikhina L.V. Toropov B.A. Makarov V.F. Ovchinsky A.S.

DOI: 10.26102/2310-6018/2024.47.4.035

The article presents a mathematical formalization of the conflict interaction of active agents focused on achieving their local goals in the process of achieving the common goal of the organizational system. The conflict is considered as a clash of active agents over a single resource, the possession of which will allow achieving a local goal. Three types of relations of an active agent to a given resource (possession, non-distinction, opposition) are presented, taking into account their usefulness in achieving a local goal. Mathematically, the conflict between agents is determined by the establishment of links between the elements of the set of active agents with the elements of the set of resources that caused the conflict. An algorithm for evaluating the mutual impact of active agents due to a resource in the core of the conflict is proposed, based on the construction of a bipartite graph "active agent - resource" and a graph of conflict in the organizational system. The weights of the arcs of a bipartite graph are defined as the values of the utility functions of the resource that caused the conflict in achieving local goals by active agents. The implementation of the algorithm allows to obtain an assessment of the degree of collision of active agents due to a single resource and an assessment of the interaction of active agents in the core of the conflict. An example of the algorithm execution is given.

Keywords: agent, conflict, resource, conflict core, local target, graph, graph weight matrix, organizational system

Using neural networks to determine dust pollution near open-pit coal mining areas based on Earth remote sensing data

2025. T.13. № 1. id 1756

Ozaryan Y.A. Kozhevnikova T.V. Tsygulev K.S. Okladnikov V.Y.

DOI: 10.26102/2310-6018/2025.48.1.007

The article examines the use of neural networks for detecting dust pollution near open-pit coal mining areas based on remote sensing data. The study involved coal mining sites located in various regions of the Russian Federation. Satellite images from the Sentinel-2 mission served as the primary data source and were processed using Quantum GIS software. An algorithm for forming the training dataset was developed, utilizing the visible and near-infrared spectral bands from the satellite imagery. The mask creation technology in the developed algorithm is based on the Enhanced Coal Dust Index and its subsequent clustering. U-Net is used as a neural network model. The trained model was tested on a validation dataset. The recognition accuracy was 59.3% for the Intersection over Union metric, 78.9% for the Precision metric, 80.6% for the F1 metric, and 95.5% for the Accuracy metric. This level of accuracy is attributed to the limited volume of training data. The potential for improving accuracy through increasing the sample size in conjunction with optimizing the parameters of the neural network is discussed. The results obtained provide a basis for assessing the environmental impacts of coal mining activities and for developing measures to ensure ecological safety based on these findings.

Keywords: dust pollution, earth remote sensing, machine learning, clustering, neural network

Model of stochastic electrical load in the residential sector using the Weibull probability density function

2024. T.12. № 4. id 1755

Borovskiy A.V. Yumenchuk A.A.

DOI: 10.26102/2310-6018/2024.47.4.034

The article proposes a method for simulating daily schedules of electrical loads in the residential sector based on convolution theory. The authors consider models using the Weibull probability density and the probability density of the normal distribution for shifts in the time of switching on household appliances. The goal is to select a model, the results of which most accurately correspond to the actual energy consumption in the residential sector. First, the energy consumption of household appliances is considered, and the results are compared without a shift and with a shift in the Weibull probability density. The correct variant of comparing the results of simulation modeling using the Weibull probability density with the results of modeling using the probability density of the normal distribution is determined. Next, the energy consumption of households in rural areas is considered, which takes into account the use of electric heating devices by the population. This makes it possible to carry out simulation modeling of energy consumption of settlements or their individual areas. The results are compared with real data on the energy consumption of the village. Based on the results of the work, a model was selected that most accurately reflects the real dynamics of changes in energy consumption levels in the residential sector. The reasons why the choice was made in favor of one of the models are described. Sufficient accuracy of simulation modeling using the selected model has been demonstrated.

Keywords: stochastic energy consumption models, simulation modeling, daily energy consumption schedule, weibull probability density, normal distribution

Influence of the TensorFlow library’s version on the quality of code generation from an image

2024. T.12. № 4. id 1754

Nikitin I.V.

DOI: 10.26102/2310-6018/2024.47.4.040

This study compares the efficiency of training models that implement two different approaches: complicating the original neural network architecture, or maintaining the architecture while improving the tools used in the training framework. Attempts to complicate the architecture of the solution for generating source code based on an image lead to solutions that might be difficult to support in the future. At the same time, such improvements do not use more modern tools and libraries that such systems are built upon. The relevance of the study is due to the lack of attempts to use more modern and relevant libraries. In this regard, during the experiment to compare the indicators of models of three versions of image-based source code generation systems: the original pix2code system, its complex version, and the version with a modern version of the TensorFlow library - in the process of their training, it was revealed that approaches with a complex architecture and the current TensorFlow have the same indicators, higher quality than the original pix2code. Based on the experiment, we can conclude that updating the TensorFlow library can provide an additional increase in the quality of results that the image-based source code generation system can predict.

Keywords: code generation, image, machine learning, tensorFlow, keras, domain-specific language

Integration of the processes of digital transformation of enterprises and training of specialists in the field of automation of industrial and agricultural production

2024. T.12. № 4. id 1753

Lishilin M.V. Pryakhin V.N. Karapetyan M.A.

DOI: 10.26102/2310-6018/2024.47.4.032

Digital transformation in industry and agriculture is aimed at driving intensive economic growth and revolutionizing project management at the state level. It involves changing management strategies and operational models through the integration of information systems. Key challenges include increasing the digital maturity of agricultural enterprises, involving stakeholders in the digital transformation process, improving access to technological information, and expanding the use of technologies like the Industrial Internet of Things and digital twins. To achieve strategic goals, it is crucial not only to enhance the digital maturity of enterprises but also to prepare qualified personnel. The integration of university and enterprise information systems, the use of virtual computer labs in education, and data-driven management are essential elements of successful digital transformation. A critical factor is the systematic use of tools such as virtual computer labs as a technological foundation for integrating production data into the educational process. However, further development of methodological and regulatory frameworks for university and enterprise information systems is needed. This will improve the quality of specialist training and actively involve universities in the processes of industrial digital transformation.

Keywords: digital transformation, virtual computer lab, information technology, training of personnel, knowledge management

Development and research of a software system for biometric authentication of a user based on the dynamics of a handwritten signature using fuzzy features

2024. T.12. № 4. id 1750

Anisimova E.S. Anikin I.V.

DOI: 10.26102/2310-6018/2024.47.4.027

The complexity of reliable biometric user authentication based on the dynamics of handwritten signatures is due to their high intra-class variability associated with changes in the physical and emotional state of a person, as well as writing conditions. Existing approaches do not always provide sufficient accuracy and resistance to these variations. This work is devoted to the study and development of a software system for biometric authentication using the apparatus of fuzzy set theory to improve the reliability of recognition. In this work, we proposed an original feature model of a dynamic handwritten signature, including a set of static and dynamic features, including fuzzy ones, taking into account the uncertainty and variability of handwriting. As a signature standard, we used a set of membership functions built on the basis of the components of the feature model. We proposed an architecture of the recognition system consisting of training subsystems, creating a test signature model, and making a decision on authenticity. We developed a software system that implements the proposed approach using the SciLab mathematical package and the C++ programming language. The system provides the functionality of user registration and formation of a training sample based on signatures entered using a graphic tablet, as well as recognition of test signatures. We conducted an experimental study based on the MCYT Signature 100 signature collection. During the study, we experimentally determined the optimal value of the cluster compactness degree for constructing feature membership functions that minimizes the equal error rate coefficient. The experimental results demonstrate a decrease in the equal error rate coefficient compared to known methods, which indicates the effectiveness of the proposed approach. The use of fuzzy features helps to increase the system's resistance to variations in signatures and, as a result, increase the reliability of biometric authentication in various applications that require identity verification. The results of the study can be used to improve the security of authentication systems and protect confidential information.

Keywords: biometric authentication, handwritten signature, graphic tablet, signature input dynamics, fuzzy set theory, fuzzy logic, signature model, signature standard, pressure pattern, writing rhythm

Method of preparing data on scientific publications for intelligent decision-making support in evaluating expertise of peer reviewers

2024. T.12. № 4. id 1748

Latypova V.A.

DOI: 10.26102/2310-6018/2024.47.4.026

One of the main factors in assigning a peer reviewer is his expertise on the manuscript topic (the existence of the relevant publicatios). Decision-making support, based on the usage of mining scientometric base data on scientific publications, speeds up the process of evaluating the expertise of peer reviewers and makes it less time-consuming. However, the critical point in this case is the correctness of the data on scientific publications subject to intellectual analysis. At present, researchers actively deal with the question of defining the scientometric base data correctness and means of ensuring it, conducting different procedures of cleaning within data preparation. Yet in the existing works, the specifics of the task, for which data on scientific publications are gathered, is not taken into account. To address the problem, a method of preparing data on scientific publications for intelligent decision-making support in evaluating expertise of peer reviewers, considering features associated with the need to define the semantic similarity of text of data on publications, is suggested in the paper. The method was successfully tested when preparing data on scientific publications of members of the academic journal “Systems Engineering and Information Technologies” editorial board, involving the content of their profiles in scientometric bases “RISC” and “Google Scholar”.

Keywords: data preparation, decision-making support, data mining, peer reviewer, scientific publication

Analysis of mudflow characteristics with limited data using machine learning models

2024. T.12. № 4. id 1747

Лютикова Л.А.

DOI: 10.26102/2310-6018/2024.47.4.029

In paper, a combined method for analyzing incomplete and distorted information is proposed, demonstrated by the example of mudflow forecasting. The main purpose of the study is to demonstrate the ability not only to create accurate forecasts, but also to analyze the decision-making mechanisms of the model, identifying significant parameters that affect predictions. To represent the identified sets of parameters affecting the volume of the mudflow in the form of logical rules, it was necessary to use data categorization. This made it possible to increase the reliability of models in the presence of emissions and noise, as well as to take into account non-linearities. Two approaches were used to form logical rules: the method of associative analysis and the original method of constructing a logical classifier. As a result of associative analysis, rules were identified that reflect certain patterns in the data, which, as it turned out, required significant correction. The use of a logical classifier made it possible to clarify and correct the patterns, ensuring the determination of a set of factors influencing the volume of mudflow. This approach made it possible to identify the most significant input variables and understand how the model processes data to generate a forecast, identify factors that play a key role in forecasting results, and ensure adequate accuracy and stability of forecasts, taking into account the specifics and complexity of mudflow data. The patterns deduced as a result of the study, reflecting the hidden principles of the subject area under study, and the methods of logical analysis used in the study helped to identify possible causes of the formation of different volumes of carried-out solid deposits. The results obtained can be used to improve monitoring systems and prevent the negative consequences of mudslides.

Keywords: machine learning, neural networks, cluster analysis, associative rules, mudflows, model

Organization of radio communication with remote mobile ground objects

2024. T.12. № 4. id 1743

Dorokhin S.V. Ivannikov V.A. Zhaivoronok D.A.

DOI: 10.26102/2310-6018/2024.47.4.028

The article considers the issues of improving the quality of information transmission on mobile objects by using modern equipment of the digital radio system DMR (Digital Mobile Radio) technology, corresponding to modern requirements for noise immunity, communication range, security of data transmission and reception. The equipment has all the advantages of digital technologies compared to analog ones, uses one channel with a frequency band of 12.5 kHz, divided in time into two logical channels. This allows you to work through a repeater with support for dual-frequency simplex technology with duplex diversity, in this mode two simultaneous independent voice connections are possible [2]. The structural diagrams of the radio interface of the proposed standard, its main advantages, characteristics, and advantages over currently used digital and analog radio systems are described. Structural schemes for the organization of communication between several subscribers have been developed, providing the possibility of simultaneous operation of two groups of users through one or more repeaters on the same channel. In order to effectively use the available data exchange resource, modern methods of channel multiplexing and their combinations are proposed. Statistical and time multiplexing using discrete multi-tone modulation allows minimizing the effects of signal attenuation with increasing frequency. The proposed technical solutions provide the possibility of gradual replacement of obsolete equipment due to the simultaneous use of analog and digital equipment, as well as effective use of the frequency range in conditions of its limited distribution.

Keywords: information transmission, system, equipment, standards, communication channels, radio signal, interference

Implementing dynamic pricing across multiple pricing groups in real estate

2025. T.13. № 1. id 1741

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

DOI: 10.26102/2310-6018/2025.48.1.009

The article presents a mathematical model of dynamic pricing for real estate that incorporates multiple pricing groups, thereby expanding the capabilities of existing models. The developed model solves the problem of maximizing aggregate cumulative revenue at the end of the sales period while meeting the revenue and sales goals. The basic formulation of the problem of optimizing revenue from the sale of all real estate objects in inventory by the end of the sales period is considered. Theoretical results describing the general form of the pricing policy for this problem are presented. A method is proposed for distributing aggregate cumulative revenue goals across different for real estate pricing groups. The model is further modified to account for the time value of money and the real estate value increase as construction progresses. The algorithm for constructing a pricing policy for multiple pricing groups is described, and numerical simulations are performed to demonstrate how the algorithm operates. The relevance of the developed model lies in the need to account for multiple pricing groups when forming the pricing policy, as well as the time value of money and the value of real estate increase as construction progresses. The obtained results can be applied to price management of real estate objects in practice.

Keywords: dynamic pricing, real estate, pricing groups, revenue maximisation, even inventory absorption, value of money, real estate value

Concept and architecture of parsing and storing a unified database of patents and scientific journal publications

2024. T.12. № 4. id 1740

Kozina S.A. Kulinchenko I.A. Korobkin D.M. Fomenkov S.A.

DOI: 10.26102/2310-6018/2024.47.4.024

The currently existing methods of automated data collection, although they facilitate this process, often face problems of low reliability, efficiency and speed. Unstable connections, blocking IP addresses and changes in the structure of sites lead to data loss and the need for constant monitoring of the parsing process, which increases the cost of maintaining and operating such systems. In this regard, the development of new approaches and tools for parsing the necessary information is a very urgent task that can transform the field of data mining. The article discusses the process of developing a module for parsing information from patent systems and websites of physics and technology journals using modern technologies and approaches, and also presents the results of checking its operability. This tool can be useful for patent offices, researchers, students, engineers, and scientists working in the subject area under consideration. The use of such a module will open up new opportunities for data mining and strategic decision-making in the field of innovative development, as well as for in-depth analysis of technological trends, identification of promising developments and building innovative development strategies.

Keywords: patents, physics and technology journals, parsing, scalability, fault tolerance

The development of a phishing attack protection system using software-hardware implementation of machine learning methods

2024. T.12. № 4. id 1738

Lukmanova K.A. Kartak V.M.

DOI: 10.26102/2310-6018/2024.47.4.033

Due to the constant evolution of phishing attacks, traditional protection methods, such as URL filtering and user training, have become insufficiently effective. The article examines modern methods of detecting phishing attacks using machine learning algorithms aimed at improving the accuracy and efficiency of URL classification. The developed system employs a multilayer perceptron for automatic URL analysis and classification of links as either phishing or legitimate. Creating a high-quality, representative dataset containing both phishing and legitimate links is one of the key stages in model development. The focus is on analyzing URL addresses based on 30 key features, including URL length, SSL certificate presence, and IP address usage. The model testing results demonstrated high accuracy, significantly surpassing the results of traditional filtering methods. The developed software, implemented in Python with TensorFlow and Scikit-Learn libraries, proved highly effective in real-world conditions, ensuring high accuracy, recall, and F1 score. The results confirm that machine learning enhances the efficiency and accuracy of phishing detection compared to traditional methods.

Keywords: phishing, cybersecurity, machine learning, multilayer perceptron, random forest, URL classification, phishing detection, data protection

Features of deep learning application for license plate detection in images and their subsequent classification using computer vision methods

2024. T.12. № 4. id 1736

Revera V.S. Shelmina E.A.

DOI: 10.26102/2310-6018/2024.47.4.042

The article presents a technique for recognizing Russian car license plates using modern technologies of deep learning, computer vision and optical character recognition. The relevance of the study is due to the growing need for automated license plate recognition systems to improve road safety, optimize traffic flows and implement intelligent transport systems. The study consists of two stages. At the first stage, a neural network was trained to detect license plates in the image using the appropriate dataset of license plate. At the second stage, based on the received detections, image processing is carried out using computer vision methods, the selection of individual characters by segmentation, as well as their subsequent classification using an optical character recognition system with an adapted alphabet. The results obtained demonstrate the effectiveness of the proposed approach and the possibility of its application in real conditions. The materials of the article are of practical value for specialists involved in the development of automatic license plate recognition systems and can be used in the areas of access control, transport monitoring and road safety.

Keywords: YOLO, license plate recognition, segmentation, object detection, optical character recognition, neural networks, computer vision, dataset

A model of inventory management in construction projects based on Markov random processes

2024. T.12. № 4. id 1734

Serebryakova E.A.

DOI: 10.26102/2310-6018/2024.47.4.025

The article describes the developed mathematical model that allows, based on probabilistic methods, to dynamically assess the possibility of risks associated with resource shortages during construction projects. The proposed model makes it possible to monitor the reserves of scarce resources during construction projects, taking into account the stochastic nature of their replenishment and expenditure, which will allow for preventive measures aimed at maintaining existing reserves at the required level. The mathematical basis of the model is the theory of stationary Markov random processes and mass service. To implement the computational procedures for the model in practice, a methodology for performing calculations in the MS Excel spreadsheet is described. Calculation using the Excel calculation sheet involves entering the required resource for which inventory management is required for each time period, as well as the range of consumption of this resource during construction work. Next, the reduced intensity of resource consumption for several periods of project implementation is calculated, and the probabilities of resource availability as a function of time and the amount of delivered resource stock. Based on the calculation sheet, it is possible to estimate the probabilities of the resource available in stock for several time forecast periods in order to minimize possible risks associated with resource shortages. The final part of the work presents the results of testing the model in the activities of a construction organization, which showed that the use of the model will save from 5 to 10% of funds associated with the costs of suspending construction work due to a lack of necessary resources.

Keywords: inventory management, construction, markov random processes, queuing theory, mathematical modeling

Mathematical modeling of the process of formation shrinkage cavity in end slabs during casting on continuous casting machine

2024. T.12. № 4. id 1732

Kabakov Z.K. Gabelaja D.I. Chuev A.A.

DOI: 10.26102/2310-6018/2024.47.4.021

The modeling of the dynamics of the profile formation and the depth of the shrinkage cavity is performed based on mathematical models of the process of volumetric and linear shrinkage, constructed using the finite difference method with an explicit scheme of approximation of partial derivatives. These models, unlike the known previous ones, take into account the different nature of the metal solidification process depending on the chemical composition and use a two-dimensional computational domain divided into a given number of nodes along the x and z coordinates. The modeling uses a system of algorithms for calculating the solidification dynamics of a continuously cast billet, linear and volumetric shrinkage of metal, as well as the process of formation of a shrinkage cavity. In addition, the influence of the carbon concentration in steel, its thermophysical properties and technological parameters of continuous casting on the process of metal solidification is taken into account. The implementation is presented in the form of a computer program, the input parameters of the modeling are the chemical composition of steel and the technological parameters of casting, the output parameters are the values of the thermophysical coefficients and the profile of the shrinkage cavity in the final slab. Verification was performed by comparing the calculated data with the experimental data and showed that the calculated data differ from the experimental data by less than 1%. The possibility of increasing the accuracy of the results by increasing the number of nodes in the thickness and height of the workpiece is shown, while the dependence of the accuracy on the number of nodes in the thickness of the workpiece is most pronounced. The proposed model allows reducing metal losses during casting associated with the formation of a shrinkage cavity and increasing the energy and resource efficiency of modern metallurgical enterprises.

Keywords: continuous casting, mathematical model, finite difference method, shrinkage cavity, end slab, approximation

Development of a multi-class classifier algorithm for a federated learning system operating in case of incomplete classes of local classifiers

2024. T.12. № 4. id 1731

Mikhalev P.A. Kutsakin M.A. Karamyhova O.V.

DOI: 10.26102/2310-6018/2024.47.4.023

The relevance of research is due the need to solve the problem of training multi-class classifier models used in federated machine learning system structure operating with a training data set that contains both publicly available data and confidential data that forming hidden classes. A similar problem arises in the context of training a classifier using a training data set, some of which consists of personal information or data of varying degrees of confidentiality. In this regard, this article is aimed at researching the features of the Gaussian mixture model of distributions as a way of representing hidden classes representing confidential data, as well as justifying the choice of an algorithmic method for finding maximum likelihood estimates of its parameters. The main method for solving the problem of identifying the parameters of hidden classes is a reasonably chosen two-stage iterative expectation-maximization procedure (EM-algorithm), which ensures strengthening the relationship between missing (confidential) data and unknown parameters of the data model represented by a Gaussian mixture of distributions. The article presents a diagram of the developed algorithm of a multi-class classifier for federated machine learning system, represented by parallel cycles of forming local learning models and their ensemble into a global learning model.

Keywords: federated machine learning, multi-class classification, confidential training data, gaussian mixture model of distributions, EM-algorithm

Building predictive agent models based on the inclusion of machine learning models in determining the state of agents

2024. T.12. № 4. id 1728

Lisovenko A.S. Tarasov D.A. Shishmareva A.S.

DOI: 10.26102/2310-6018/2024.47.4.017

The field of agent modeling continues to evolve towards the creation of more intelligent agents. This raises the conceptual problem of finding a balance between the determinism of agents' behavior and the ability of these agents to learn and predict their condition. One of the potential ways to solve this problem is to consider the possibility of developing an intermediate approach in the creation of agents, in which agents maintain the determinism of their behavior, but at the same time are able to predict their condition and correct behavior. The article presents a new approach to building intelligent agents, which combines the classical approach of building agents based on a priori set rules and the application of machine learning methods in the rules of agent behavior. A mathematical description of the proposed function for calculating the state of an agent using machine learning models is presented, as well as an algorithm for calculating the states of agents in the model. An example of building an agent model using the proposed approach is also given. The proposed approach makes it possible to develop agent models of complex systems in which agents are reactive but are able to predict their state and take into account the predict in determining their current state.

Keywords: agent modeling, intelligent agents, the approach of building intelligent agents, predicting the state, machine learning

Optimization of management in an organizational system with alternative supplies based on a multivariate choice model and algorithm

2024. T.12. № 4. id 1727

Lvovich Y.E. Shevyreva E.A.

DOI: 10.26102/2310-6018/2024.47.4.018

The article proposes an optimization approach to decision-making in management in an organizational system with alternative supplies based on a multivariate choice model and algorithm. The main features that define the structure of optimization model of multivariate choice are characterized: multicriteria, individual need for deliveries for each nomenclature unit and each object of the organizational system, the choice of supplier. It is shown that the initial model is a multi-criteria optimization model in which the criteria are specified on a set of alternative variables. An equivalent approach to the optimization problem with a constraint on the total supply goal for each object and the target function in the form of a weighted average convolution of the other indicators affecting the performance of the organizational system is justified. For the subsequent algorithmic multivariate selection, the target function and constraints are combined with an additive function to which extreme requirement of max-min is imposed. The algorithmic procedure of multivariate selection of management decisions is formulated by integrating a randomized search based on the task of multi-alternative optimization and genetic algorithm. The advantage in terms of search labor intensity when combining the used algorithms in the alternation mode in comparison with the known use of genetic algorithm only at the final stage of selecting the final management decision on the set of dominating options is shown.

Keywords: management, organizational system, alternative supplies, optimization, randomized search, genetic algorithm

Development of adaptive exponential min sum decoding algorithm

2024. T.12. № 4. id 1725

Zhang W. Mouhamad I. Saklakov V.M.

DOI: 10.26102/2310-6018/2024.47.4.019

This paper presents an optimized min sum (MS) decoding algorithm with low complexity and high decoding performance for LDPC short codes. The MS algorithm has low computational complexity and is simple to deploy. The MS decoding algorithm, while demonstrating a performance gap compared to the belief propagation (BP) and likelihood ratio BP (LLR-BP) decoding algorithms, shows significant potential for optimization. To improve the decoding performance of traditional MS algorithm, secondary external information is introduced into the control node (CNs) update operations of MS algorithm and optimized as adaptive exponential correction factor (AECF). The optimized MS algorithm is named as adaptive exponential exponential MS decoding algorithm (AEMS). The decoding efficiency of the AEMS algorithm for regular, irregular and LDPC codes of the Consultative Committee on Space Data Systems (CCSDS) was extensively tested, then the complexity of the AEMS algorithm was analyzed and compared with other decoding algorithms. The results show that the AEMS algorithm outperforms the offset MS (OMS) and normalized MS (NMS) algorithms in decoding performance, and outperforms the BP algorithm as the signal-to-noise ratio (SNR) gradually increases.

Keywords: LDPC, adaptive exponential algorithm, min sum, low complexity, LLR-BP

Architecture of the patient's health predicting system using agent-based modeling and machine learning methods

2024. T.12. № 4. id 1724

Lisovenko A.S. Limanovskaya O.V. Tarasov D.A. Meshchaninov V.N. Gavrilov I.V.

DOI: 10.26102/2310-6018/2024.47.4.016

The introduction of information technology in medical institutions contributes to the development of predictive, preventive and personalized medicine. The task that arises in this case is to create a software analogue of the patient, capable of taking into account his individual indicators and predicting the state of health, is still relevant. The architecture of the patient's health predicting system presented in the work is aimed at solving this problem. A distinctive feature of the system architecture is the combination of the principles of agent modeling and representation of the patient's body in the form of interacting modules, which opens up wide opportunities for modeling the health status of the patient's body. The paper describes the hierarchy of agents in the system architecture, describes the rules of agent interaction and provides a mathematical model for evaluating the effectiveness of therapeutic effects on the patient's body, the solution of which is achieved through the interaction of system agents. The prediction of the patient's health status is performed using downloadable pre-trained machine learning models, while the models are directly involved in determining the behavior of agents. The architecture of the patient's health predicting system, implemented in the form of a software package, is a powerful tool for building agent-based predicting models aimed at modeling physiological and pathological processes and effects on the patient's body.

Keywords: health predicting, patient's health, agent-based modeling, patient's digital double, modular approach

Characteristic function of an actor computing system

2024. T.12. № 4. id 1722

Zekenskii A.A. Gribkov A.A.

DOI: 10.26102/2310-6018/2024.47.4.015

The paper is devoted to the study of the problem of determining a complex feasibility indicator for an actor computing system, which can be expressed as a binary characteristic function. This function depends on the solvability and enumerability of the set of intermediate values of the parameters of the computational problem to be solved, the feasibility of the computational system, i.e. its ability to perform the entire set of necessary computational operations for a given limited time interval (computation cycle), as well as on the degree of confidence in the functional reliability and information security of the computational system, expressed in the form of an integral confidence index. The paper presents a description of the actor model of a computing system in the framework of number theory. The proposed description is based on the representation of a computing system in the form of a composition of actors – function carriers, definitions of computability of these functions, as well as solvability and enumerability of numerical sets of parameter values set for a computing system and arising in it in the process of solving the set tasks. Approaches to ensuring solvability, realisability and trust in the computational system are considered. It is stated that the choice of memory-oriented architecture of computations based on the requirement of realisability is also reasonable from the point of view of providing decidability, enumerability and ensuring trust to the computing system.

Keywords: computing system, actor model, memory-oriented architecture, feasibility, realisability, computability, solvability, enumerability, confidence

Topological skeletons-based correction of volumetric motion artifacts in optical coherence tomography and elastography

2024. T.12. № 4. id 1715

Potlov A.Y.

DOI: 10.26102/2310-6018/2024.47.4.008

An original approach to image stabilization in optical coherence tomography and elastography was presented. The key features of the proposed approach are: I) binarization and application of mathematical morphology digital operations; II) parallel construction of a topological skeleton for each optical image with an emphasis on the equivalent high- and low-level signal; III) complexing of topological skeletons; IV) comparison of a sequence of optical images by combined topological skeletons using «quench» points; V) compensation of volumetric motion artifacts by «reassembling» the original sets of interference signals. The computational efficiency of the proposed method with respect to the dynamics of interference signal acquisition by a specific device was achieved by using sequential and parallel operations. Сomputations using the central and graphical processing units, namely GPU and CPU, were combined for this. High efficiency of volumetric motion artifact correction in optical coherence tomography and elastography is ensured by robustness of topological skeletons constructed with emphasis on high-level equivalent signal to speckle noise corresponding to constructive interference (bright speckles). Topological skeletons for low-level equivalent signal are correspondingly robust to dark speckles (destructive interference result).

Keywords: optical coherence tomography, medical elastography, fiber optic probe, structural images, functional images, topological skeleton, biological tissue, tissue-imitating phantoms, volumetric motion artifacts

Improving the objectivity of the decision-making process in relation to the tasks of sowing crops

2024. T.12. № 4. id 1714

Strygin S.P. Konkina V.V. Pustovarov N.Y. Konovalcev A.S. Dudin M.A.

DOI: 10.26102/2310-6018/2024.47.4.012

The relevance of the study stems from the need to improve the efficiency and economic benefits of crop cultivation. The research in this paper is aimed at developing a decision support system that will improve the method of evaluating the accuracy of seeding units, allow pre-sowing adjustment of row crop seeders and reduce the workload of agronomists. The dotted-nested sowing method was considered, and the total coefficient of variation, dynamic coefficient of variation and accuracy were determined as criteria for assessing the unevenness of seed distribution in the row. As alternatives, soybean varieties “Alaska” and “Lisbon” of different fractions were studied at different design and operating parameters of the seeding unit, namely: the rotation speed of the seeding disk of the unit (15–55 rpm), the position of the seed wiper (from fully open hole to overlapped by 1/3 of the area of the hole), the diameter of the holes of the seeding disk (4–4.5 mm). The paper formulates the problem of decision-making theory within the framework of a specific research area. The problem is solved using the method of hierarchical analysis and complete enumeration. The article's materials are of practical use for agricultural enterprises, including pre-sowing rowed seeder settings.

Keywords: multicriteria problem, decision making, method of hierarchy analysis, weight coefficients, objective choice