Volodymyr Donets
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Volodymyr Donets has a total of 15 co-authors. Here are the top co-authors by number of shared publications:
- Viktoriia Strilets
V. N. Karazin Kharkiv National University, Ukraine
7 shared publications - Dmytro Shevchenko
V. N. Karazin Kharkiv National University
3 shared publications - Grygoriy Zholtkevych
V. N. Karazin Kharkiv National University
3 shared publications - Maksym Holikov
V. N. Karazin Kharkiv National University, Ukraine
3 shared publications - Mykhaylo Ugryumov
V. N. Karazin Kharkiv National University, Kharkiv
3 shared publications - Oleksandr Deineha
V. N. Karazin Kharkiv National University
3 shared publications
Latest Publications
- 7 February 2026
- Published by Academy of Cognitive and Natural Sciences
- p. 94-101
Analysis of the Influence of Cosine Distance Threshold Values in a Real-time Face Recognition System
The quality of a face recognition system largely depends on the correct choice of threshold value when comparing vector features (embeddings). This paper investigates the impact of cosine distance thresholds on the performance of a real-time face recognition system. The proposed approach combines face detection using MediaPipe FaceMesh and feature extraction using the ArcFace model. A series of experiments with different threshold values was conducted, the results of which were evaluated using the following metrics: Accuracy, Precision, Recall, False Accept Rate (FAR), and False Reject Rate (FRR). The results show that the choice of threshold directly determines the trade-off between security and convenience of the system. The optimal range of cosine distance threshold values was established as 0.05– 0.07, which minimizes both FAR and FRR, which is important for practical use in video surveillance, access control, and user authentication systems.
- 7 February 2026
- Published by Academy of Cognitive and Natural Sciences
- p. 94-101
Analysis of the Influence of Cosine Distance Threshold Values in a Real-time Face Recognition System
The quality of a face recognition system largely depends on the correct choice of threshold value when comparing vector features (embeddings). This paper investigates the impact of cosine distance thresholds on the performance of a real-time face recognition system. The proposed approach combines face detection using MediaPipe FaceMesh and feature extraction using the ArcFace model. A series of experiments with different threshold values was conducted, the results of which were evaluated using the following metrics: Accuracy, Precision, Recall, False Accept Rate (FAR), and False Reject Rate (FRR). The results show that the choice of threshold directly determines the trade-off between security and convenience of the system. The optimal range of cosine distance threshold values was established as 0.05– 0.07, which minimizes both FAR and FRR, which is important for practical use in video surveillance, access control, and user authentication systems.
- Open Access
- 11 July 2025
- Published by National Technical University Kharkiv Polytechnic Institute in Bulletin of National Technical University "KhPI". Series: System Analysis, Control and Information Technologies
- No. 1 (13),p. 83-87
INTELLIGENT ANALYSIS OF OPTICAL IMAGES BASED ON A HYBRID APPROACH
The article considers an intelligent approach to real-time analysis of optical images based on a combination of face recognition methods using deep learning and classical computer vision algorithms for tracking them. A system with hybrid approach is proposed that integrates preliminary face recognition based on vector features (embeddings) generated by the FaceNet neural network and face tracking using the CSRT (Channel and Spatial Reliability Tracker) algorithm, which is part of the OpenCV library. The implemented system allows to recognise and automatically identify users in a video stream from a webcam, store new faces in the database, and effectively track identified faces over subsequent frames. The frame processing algorithm is implemented in a multi-threaded mode using queues and thread synchronisation mechanisms to ensure stable operation in real time. To recognise unknown persons, a unique ID is automatically created and their features are added to the common database of emblems. Particular attention is paid to assessing the spatial overlap of detection zones to avoid duplication of trackers when several people are present in the frame at the same time. In addition, the system is implemented as a web service based on Flask, which provides convenient integration with other software modules and the possibility of remote monitoring via a web interface. The proposed hybrid approach combines the accuracy of modern deep learning models with the flexibility of classical tracking algorithms, making the system suitable for use in security systems, smart offices, educational environments, and other areas where accurate face identification in dynamic environments is important. In summary, this paper demonstrates the practical implementation of an intelligent image analysis system that can be adapted to various use cases, including video surveillance, access control, and crowd management systems, as well as research and educational projects.
- 11 July 2025
- Published by National Technical University Kharkiv Polytechnic Institute in Bulletin of National Technical University "KhPI". Series: System Analysis, Control and Information Technologies
- No. 1 (13),p. 83-87
Open AccessINTELLIGENT ANALYSIS OF OPTICAL IMAGES BASED ON A HYBRID APPROACH
The article considers an intelligent approach to real-time analysis of optical images based on a combination of face recognition methods using deep learning and classical computer vision algorithms for tracking them. A system with hybrid approach is proposed that integrates preliminary face recognition based on vector features (embeddings) generated by the FaceNet neural network and face tracking using the CSRT (Channel and Spatial Reliability Tracker) algorithm, which is part of the OpenCV library. The implemented system allows to recognise and automatically identify users in a video stream from a webcam, store new faces in the database, and effectively track identified faces over subsequent frames. The frame processing algorithm is implemented in a multi-threaded mode using queues and thread synchronisation mechanisms to ensure stable operation in real time. To recognise unknown persons, a unique ID is automatically created and their features are added to the common database of emblems. Particular attention is paid to assessing the spatial overlap of detection zones to avoid duplication of trackers when several people are present in the frame at the same time. In addition, the system is implemented as a web service based on Flask, which provides convenient integration with other software modules and the possibility of remote monitoring via a web interface. The proposed hybrid approach combines the accuracy of modern deep learning models with the flexibility of classical tracking algorithms, making the system suitable for use in security systems, smart offices, educational environments, and other areas where accurate face identification in dynamic environments is important. In summary, this paper demonstrates the practical implementation of an intelligent image analysis system that can be adapted to various use cases, including video surveillance, access control, and crowd management systems, as well as research and educational projects.
- Open Access
- 28 June 2024
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 3 (2 (129)) , 42-54
The approach development of data extraction from lambda terms
The study's object is the process of extracting the characteristics of lambda terms, which indicate the optimality of the reduction strategy and increase the productivity of compilers and interpreters. The solution to the problem of extracting specific strategy priority data from lambda terms using Machine Learning methods was considered. Such data was extracted using the large language model Microsoft CodeBERT, which was trained to solve the problem of summarizing the software code. The resulting matrices of embeddings were used to obtain vectors of average embeddings of size 768 and a latent space of size 8 thousand. Further, vectors of average embeddings were used for cluster analysis using the DBSCAN and Hierarchical Agglomerative clustering methods. The most informative variables affecting clustering were determined. Next, the clustering results were compared with the priorities of reduction strategies, which showed the impossibility of separating terms with RI priority. A feature of the obtained results is using machine learning methods to obtain knowledge. The clustering results showed many of the same informative variables, which is explained by the similar shape of the obtained clusters. The results of comparing the clustering values with the real priority are explained by the impossibility of clearly determining the priority and the use of the Microsoft CodeBERT model, which was not trained for the analysis of lambda terms. The proposed approach can find application in the development of compilers and interpreters of functional programming languages, allowing to analyze the code and extract important data to optimize the execution of programs. The obtained data can be used to develop rules aimed at improving the efficiency of compilation and interpretation
- 28 June 2024
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 3 (2 (129)) , 42-54
Open AccessThe approach development of data extraction from lambda terms
The study's object is the process of extracting the characteristics of lambda terms, which indicate the optimality of the reduction strategy and increase the productivity of compilers and interpreters. The solution to the problem of extracting specific strategy priority data from lambda terms using Machine Learning methods was considered. Such data was extracted using the large language model Microsoft CodeBERT, which was trained to solve the problem of summarizing the software code. The resulting matrices of embeddings were used to obtain vectors of average embeddings of size 768 and a latent space of size 8 thousand. Further, vectors of average embeddings were used for cluster analysis using the DBSCAN and Hierarchical Agglomerative clustering methods. The most informative variables affecting clustering were determined. Next, the clustering results were compared with the priorities of reduction strategies, which showed the impossibility of separating terms with RI priority. A feature of the obtained results is using machine learning methods to obtain knowledge. The clustering results showed many of the same informative variables, which is explained by the similar shape of the obtained clusters. The results of comparing the clustering values with the real priority are explained by the impossibility of clearly determining the priority and the use of the Microsoft CodeBERT model, which was not trained for the analysis of lambda terms. The proposed approach can find application in the development of compilers and interpreters of functional programming languages, allowing to analyze the code and extract important data to optimize the execution of programs. The obtained data can be used to develop rules aimed at improving the efficiency of compilation and interpretation
- Open Access
- 30 April 2024
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 2 (9 (128)) , 6-16
Application of a data stratification approach in computer medical monitoring systems
The research object is the processes occurring in the data stratification subsystem in the medical monitoring computer system, which is part of the decision support system. Such a subsystem aims to solve data analysis and processing problems in the medical monitoring system. Among them, the problems of anomaly detection, data marking, state determination, selection of the most informative variables, and justification of decision-making are selected for solving. The paper proposes the structure and implementation of the data stratification subsystem in the decision support system. The subsystem contains modules for anomaly detection and an autoencoder, a clustering module using an advanced multi-agent clustering method, and a state detection module with a modified neural network training procedure. Modules of the stratification subsystem have been tested using diabetes monitoring data. The results showed that the clustering module provides 25.7 % lower accuracy than the achieved neural network. The accuracy difference is explained by the complexity of the data and the lack of adaptability of the proposed method to solving such problems. It is shown that the method of determining the overall informativeness of variables covers 90 % informativeness with 10 variables, comparable to the variability data. In general, the flexible nature of the proposed stratification subsystem allows for solving the problems. The proposed stratification subsystem offers a robust solution for improving treatment strategies and decision-making in computer medical monitoring systems. Its versatility allows it to be used in any system where decision support is needed, providing valuable information about informative variables and decision-making features for clinicians and researchers
- 30 April 2024
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 2 (9 (128)) , 6-16
Open AccessApplication of a data stratification approach in computer medical monitoring systems
The research object is the processes occurring in the data stratification subsystem in the medical monitoring computer system, which is part of the decision support system. Such a subsystem aims to solve data analysis and processing problems in the medical monitoring system. Among them, the problems of anomaly detection, data marking, state determination, selection of the most informative variables, and justification of decision-making are selected for solving. The paper proposes the structure and implementation of the data stratification subsystem in the decision support system. The subsystem contains modules for anomaly detection and an autoencoder, a clustering module using an advanced multi-agent clustering method, and a state detection module with a modified neural network training procedure. Modules of the stratification subsystem have been tested using diabetes monitoring data. The results showed that the clustering module provides 25.7 % lower accuracy than the achieved neural network. The accuracy difference is explained by the complexity of the data and the lack of adaptability of the proposed method to solving such problems. It is shown that the method of determining the overall informativeness of variables covers 90 % informativeness with 10 variables, comparable to the variability data. In general, the flexible nature of the proposed stratification subsystem allows for solving the problems. The proposed stratification subsystem offers a robust solution for improving treatment strategies and decision-making in computer medical monitoring systems. Its versatility allows it to be used in any system where decision support is needed, providing valuable information about informative variables and decision-making features for clinicians and researchers
- 26 December 2023
- Published by Igor Sikorsky Kyiv Polytechnic Institute in System research and information technologies
- No. 4,p. 21-36
Methodology of the countries’ economic development data analysis
The paper examines the issue of improving the methods of identification of economic objects and their analysis using algorithms of intelligent data processing. The use of the developed methodology in the economic analysis allows for improvement in the quality of management. It can be the basis for creating decision support systems to prevent potentially dangerous changes in the economic status of the research object. In this work, an improved method of c-means data clustering with agent-oriented modification is proposed, and a radial-basis neural network and its extension are proposed to determine whether the obtained clusters are relevant and to analyze the informativeness of state variables and obtain a subset of informative variables. The effect of applying data compression using an autoencoder on the accuracy of the methods is also considered. According to the results of testing of the developed methodology, it was proved that the probability of incorrect determination of the state was reduced when identifying the states of economic systems, and a reduced value of the error of the third kind was obtained when classifying the states of objects.
- 26 December 2023
- Published by Igor Sikorsky Kyiv Polytechnic Institute in System research and information technologies
- No. 4,p. 21-36
Methodology of the countries’ economic development data analysis
The paper examines the issue of improving the methods of identification of economic objects and their analysis using algorithms of intelligent data processing. The use of the developed methodology in the economic analysis allows for improvement in the quality of management. It can be the basis for creating decision support systems to prevent potentially dangerous changes in the economic status of the research object. In this work, an improved method of c-means data clustering with agent-oriented modification is proposed, and a radial-basis neural network and its extension are proposed to determine whether the obtained clusters are relevant and to analyze the informativeness of state variables and obtain a subset of informative variables. The effect of applying data compression using an autoencoder on the accuracy of the methods is also considered. According to the results of testing of the developed methodology, it was proved that the probability of incorrect determination of the state was reduced when identifying the states of economic systems, and a reduced value of the error of the third kind was obtained when classifying the states of objects.
- 1 December 2023
- Published by Springer Nature
- p. 25-38
On Randomization of Reduction Strategies for Typeless Lambda Calculus
- 1 December 2023
- Published by Springer Nature
- p. 25-38
On Randomization of Reduction Strategies for Typeless Lambda Calculus
- Open Access
- 31 October 2022
- Published by V. N. Karazin Kharkiv National University in Bulletin of V.N. Karazin Kharkiv National University, series «Mathematical modeling. Information technology. Automated control systems»
- Vol. 55 (55) , 6-18
Agent-oriented method of clustering the wholesale distributor data
The purpose of the research is to improve the accuracy of data clustering and to determine the target number of data clusters generated by dynamic economic systems, using an agent-oriented clustering method with the introduction of data preprocessing methods. Research methods: data processing and preparation methods, elemental distance measures, and clustering methods have been used. The software is developed by using the Python language. The following libraries have also been used: scikit-learn, NumPy, SciPy, Pandas, PyTorch and others. As a result of the research, the data of the wholesale distributor have been processed by the data pre-processing methods such as the determination of missing values, the determination of asymmetry and the Box-Cox transformation. The normalization of the data with the min-max normalization method and the dimensionality reduction with the PCA and t-SNE methods have been carried out. Afterwards, the agent-oriented clustering method has been applied with the Manhattan distance, Mahalanobis distance with the inverse value of the membership function, Kullback-Leibler divergence and cross-entropy metrics. Kullback-Leibler divergence has shown the best accuracy results and has been chosen for the further testing. The ability of the agent-oriented method to determine the number of clusters has been tested. The use of data preprocessing methods shows the clear presence of 3 target clusters, which was confirmed by the method. Conclusions: The developed method allows for high clustering accuracy due to the performed data processing, the correctly selected measure of elemental distance and the use of an agent-oriented approach. This method can be used to improve the quality of data clustering of dynamic economic systems, but the method requires improvement in order to increase flexibility in determining the size of cluster agents.
- 31 October 2022
- Published by V. N. Karazin Kharkiv National University in Bulletin of V.N. Karazin Kharkiv National University, series «Mathematical modeling. Information technology. Automated control systems»
- Vol. 55 (55) , 6-18
Open AccessAgent-oriented method of clustering the wholesale distributor data
The purpose of the research is to improve the accuracy of data clustering and to determine the target number of data clusters generated by dynamic economic systems, using an agent-oriented clustering method with the introduction of data preprocessing methods. Research methods: data processing and preparation methods, elemental distance measures, and clustering methods have been used. The software is developed by using the Python language. The following libraries have also been used: scikit-learn, NumPy, SciPy, Pandas, PyTorch and others. As a result of the research, the data of the wholesale distributor have been processed by the data pre-processing methods such as the determination of missing values, the determination of asymmetry and the Box-Cox transformation. The normalization of the data with the min-max normalization method and the dimensionality reduction with the PCA and t-SNE methods have been carried out. Afterwards, the agent-oriented clustering method has been applied with the Manhattan distance, Mahalanobis distance with the inverse value of the membership function, Kullback-Leibler divergence and cross-entropy metrics. Kullback-Leibler divergence has shown the best accuracy results and has been chosen for the further testing. The ability of the agent-oriented method to determine the number of clusters has been tested. The use of data preprocessing methods shows the clear presence of 3 target clusters, which was confirmed by the method. Conclusions: The developed method allows for high clustering accuracy due to the performed data processing, the correctly selected measure of elemental distance and the use of an agent-oriented approach. This method can be used to improve the quality of data clustering of dynamic economic systems, but the method requires improvement in order to increase flexibility in determining the size of cluster agents.
- 23 February 2022
- Published by National Aerospace University - Kharkiv Aviation Institute in RADIOELECTRONIC AND COMPUTER SYSTEMS
- No. 1,p. 103-114
Agent-oriented data clustering for medical monitoring
Medical data processing is one of the priority machine learning areas. Usually, data obtained in the process of medical patient monitoring are complex and have a different nature. Solving the problem of clustering, classification, or forecasting problem these data requires the creation of new methods or improvement of existing methods to improve the decision accuracy and effectiveness. The classical clustering approaches and the c-means fuzzy clustering method were analyzed. Based on the multiagent systems theory, it is proposed to use in the c-means method the separate rules for selecting elites when forming clusters and selecting the best of them in accordance with the chosen intra-cluster distance measures. The result of solving such a problem is the number of clusters, as well as the number of elements in them. The method quality was tested on Fisher iris data set using three measures of intra-cluster distance: Mahalanobis distance, Mahalanobis distance considering the membership function, and Kullbak-Leibler entropy. The highest accuracy of 98% was obtained for the distance measured by the Kullbak-Leibler entropy. Therefore, this measure was chosen to solve the clustering problem of medical monitoring data for prostate disease. Medical monitoring data were divided into four classes of patient states: “healthy persons”, “non-metastatic patients”, “metastatic patients” and “hormone-resistant patients”. The accuracy of clustering according to medical data was 95,6%. In addition to accuracy, the confusion matrix, ROC- and LF-curves were used to assess the method quality. The minimum value of the ROC-curve was 0.96 for Fisher's irises and 0.95 for medical monitoring data, which characterizes the high quality of the proposed clustering method. The loss function value is also quite small (-0.056 and -0.0176 for each considered data set), which means that the optimal cluster number and the distribution of data over them are obtained. Based on the obtained results analysis, the proposed method can be recommended for use in medical information and diagnostic decision support systems for clustering monitoring data.
- 23 February 2022
- Published by National Aerospace University - Kharkiv Aviation Institute in RADIOELECTRONIC AND COMPUTER SYSTEMS
- No. 1,p. 103-114
Agent-oriented data clustering for medical monitoring
Medical data processing is one of the priority machine learning areas. Usually, data obtained in the process of medical patient monitoring are complex and have a different nature. Solving the problem of clustering, classification, or forecasting problem these data requires the creation of new methods or improvement of existing methods to improve the decision accuracy and effectiveness. The classical clustering approaches and the c-means fuzzy clustering method were analyzed. Based on the multiagent systems theory, it is proposed to use in the c-means method the separate rules for selecting elites when forming clusters and selecting the best of them in accordance with the chosen intra-cluster distance measures. The result of solving such a problem is the number of clusters, as well as the number of elements in them. The method quality was tested on Fisher iris data set using three measures of intra-cluster distance: Mahalanobis distance, Mahalanobis distance considering the membership function, and Kullbak-Leibler entropy. The highest accuracy of 98% was obtained for the distance measured by the Kullbak-Leibler entropy. Therefore, this measure was chosen to solve the clustering problem of medical monitoring data for prostate disease. Medical monitoring data were divided into four classes of patient states: “healthy persons”, “non-metastatic patients”, “metastatic patients” and “hormone-resistant patients”. The accuracy of clustering according to medical data was 95,6%. In addition to accuracy, the confusion matrix, ROC- and LF-curves were used to assess the method quality. The minimum value of the ROC-curve was 0.96 for Fisher's irises and 0.95 for medical monitoring data, which characterizes the high quality of the proposed clustering method. The loss function value is also quite small (-0.056 and -0.0176 for each considered data set), which means that the optimal cluster number and the distribution of data over them are obtained. Based on the obtained results analysis, the proposed method can be recommended for use in medical information and diagnostic decision support systems for clustering monitoring data.
- 22 January 2022
- Published by Academy of Cognitive and Natural Sciences
Estimating Lambda-Term Reduction Complexity with Regression Methods
In this study, we dive deep into the Pure Lambda Calculus' evaluation process, specifically zooming in on term reduction steps. Commonly, these steps are seen as the same, but our research suggests otherwise - some steps are more complex than others. We investigated how specific term features play a role in computing efficiency, aiming to shed light on how different reduction strategies perform. Through testing, our Linear Regression and ANN regression models showed potential in predicting the time taken for a reduction step based on term details. Yet, when we used richer datasets to pin down complexity, our predictions didn't sharpen as much as hoped. This leads us to believe that term details, while useful, might not be the full puzzle pieces we need for spot-on strategy predictions. This research opens the door for further exploration into refining our models and unearthing other crucial factors that could shape term reduction time. Ultimately, our findings could help optimize programming tools like compilers and interpreters, steering them toward swifter operations.
- 22 January 2022
- Published by Academy of Cognitive and Natural Sciences
Estimating Lambda-Term Reduction Complexity with Regression Methods
In this study, we dive deep into the Pure Lambda Calculus' evaluation process, specifically zooming in on term reduction steps. Commonly, these steps are seen as the same, but our research suggests otherwise - some steps are more complex than others. We investigated how specific term features play a role in computing efficiency, aiming to shed light on how different reduction strategies perform. Through testing, our Linear Regression and ANN regression models showed potential in predicting the time taken for a reduction step based on term details. Yet, when we used richer datasets to pin down complexity, our predictions didn't sharpen as much as hoped. This leads us to believe that term details, while useful, might not be the full puzzle pieces we need for spot-on strategy predictions. This research opens the door for further exploration into refining our models and unearthing other crucial factors that could shape term reduction time. Ultimately, our findings could help optimize programming tools like compilers and interpreters, steering them toward swifter operations.
- 8 November 2020
- Published by Academy of Cognitive and Natural Sciences
Machine Learning Methods in Medicine Diagnostics Problem
Medical service improvement has always been a life topical problem. To decide it, we must continuously raise the competency of doctors on the one hand and it is necessary to develop new methods and approaches which could help take decisions concerning diagnostics (classification) of patient health conditions and concerning patient’s further treatment. At the paper the machine learning methods for patient health condition clas- sification were considered. These methods were Naive Bayes Classifier, Linear Classifier, Support-vector machine, K-nearest Neighbor Classifier, Logistic Re- gression, Decision Tree Classifier, Random Forest Classifier, Ada Boost Classi- fier and Artificial Neural Network. A radial basis network was chosen from the variety of artificial neural system architecture to solve classification tasks. The problem of patient health conditions classification was considered for two sets of laboratory research results: on liver diseases and on urological dis- eases. Confusion matrixes and ROC-curves were taken to estimate classification quality of patient health conditions with above-mentioned methods.
- 8 November 2020
- Published by Academy of Cognitive and Natural Sciences
Machine Learning Methods in Medicine Diagnostics Problem
Medical service improvement has always been a life topical problem. To decide it, we must continuously raise the competency of doctors on the one hand and it is necessary to develop new methods and approaches which could help take decisions concerning diagnostics (classification) of patient health conditions and concerning patient’s further treatment. At the paper the machine learning methods for patient health condition clas- sification were considered. These methods were Naive Bayes Classifier, Linear Classifier, Support-vector machine, K-nearest Neighbor Classifier, Logistic Re- gression, Decision Tree Classifier, Random Forest Classifier, Ada Boost Classi- fier and Artificial Neural Network. A radial basis network was chosen from the variety of artificial neural system architecture to solve classification tasks. The problem of patient health conditions classification was considered for two sets of laboratory research results: on liver diseases and on urological dis- eases. Confusion matrixes and ROC-curves were taken to estimate classification quality of patient health conditions with above-mentioned methods.