Ihor Bosenko
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ihor Bosenko has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Oleksii Matsiievskyi
Kyiv National University of Construction and Architecture,Dept. of Information Technologies,Kyiv,Ukraine
3 shared publications - Bohdan Volokh
Київський національний університет будівництва і архітектури, м. Київ
2 shared publications - Igor Achkasov
Kyiv National University of Construction and Architecture,Dept. of Information Technologies,Kyiv,Ukraine
2 shared publications - Roman Mazurenko
Київський національний університет будівництва і архітектури, Київ, Україна, Україна
2 shared publications - Svitlana Terenchuk
Київський національний університет будівництва і архітектури
2 shared publications - Yevhenii Biloshchytskyi
Astana IT University,Department of Computer Engineering,Astana,Kazakhstan
2 shared publications
Latest Publications
- Open Access
- 24 September 2026
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 67,p. 38-45
Automated detection of cracks in concrete structures using a deep feature extraction method based on convolutional neural networks
The study addresses the task of automating the diagnostics of the technical condition of concrete infrastructure facilities through computer-based image analysis. The aim of the work is to develop and experimentally evaluate an approach to classifying cracks on the concrete surface based on deep features extracted using a pre-trained convolutional neural network SqueezeNet1.1 without fine-tuning its weights, in comparison with classical texture descriptors. For comparative analysis, a vector of classical features was formed (HOG, GLCM statistics, edge density based on the Canny operator). All extracted features were independently normalized to align the scale of the variables. For both feature sets, the effectiveness of four machine learning algorithms (logistic regression, random forest, SVM, gradient boosting) was evaluated using 5-fold stratified cross-validation with Accuracy, Precision, Recall, and F1-score metrics. To ensure the interpretability of the decisions made, occlusion-based sensitivity analysis was applied, which does not require gradient computation. The advantage of deep features was experimentally demonstrated: the SVM model achieved an accuracy of 99.6% versus 98.3% for the best configuration on classical features. Sensitivity maps confirmed that the model's attention was localized directly on the geometry of the cracks. The use of a compact fixed feature extractor minimizes the computational resources required for model deployment. The results obtained substantiate the practical suitability of the approach for implementation in low-resource automated monitoring systems for building structures, supported by explainable artificial intelligence (XAI) methods.
- 24 September 2026
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 67,p. 38-45
Open AccessAutomated detection of cracks in concrete structures using a deep feature extraction method based on convolutional neural networks
The study addresses the task of automating the diagnostics of the technical condition of concrete infrastructure facilities through computer-based image analysis. The aim of the work is to develop and experimentally evaluate an approach to classifying cracks on the concrete surface based on deep features extracted using a pre-trained convolutional neural network SqueezeNet1.1 without fine-tuning its weights, in comparison with classical texture descriptors. For comparative analysis, a vector of classical features was formed (HOG, GLCM statistics, edge density based on the Canny operator). All extracted features were independently normalized to align the scale of the variables. For both feature sets, the effectiveness of four machine learning algorithms (logistic regression, random forest, SVM, gradient boosting) was evaluated using 5-fold stratified cross-validation with Accuracy, Precision, Recall, and F1-score metrics. To ensure the interpretability of the decisions made, occlusion-based sensitivity analysis was applied, which does not require gradient computation. The advantage of deep features was experimentally demonstrated: the SVM model achieved an accuracy of 99.6% versus 98.3% for the best configuration on classical features. Sensitivity maps confirmed that the model's attention was localized directly on the geometry of the cracks. The use of a compact fixed feature extractor minimizes the computational resources required for model deployment. The results obtained substantiate the practical suitability of the approach for implementation in low-resource automated monitoring systems for building structures, supported by explainable artificial intelligence (XAI) methods.
- Open Access
- 25 June 2026
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 66,p. 4-9
Method for generating adaptive learning recommendations based on semantic analysis of students’ responses
This study is devoted to developing a method for improving the efficiency of the educational process through the automated analysis of students’ responses and the generation of personalized recommendations. The object of the study is the process of generating adaptive learning recommendations based on the automated analysis of students’ open-ended textual responses. The subject of the study comprises models and methods of neural-network-based semantic analysis that enable the determination of the semantic correspondence between a student’s response and a reference answer, the completeness of key concept coverage, and the presence of content gaps. The aim of the study is to provide a theoretical justification and formalization of a method for generating adaptive learning recommendations based on the semantic analysis of students’ responses. A Siamese transformer neural network based on Sentence-BERT is proposed for generating vector representations of texts, enabling the consideration of context, synonymous constructions, and paraphrasing. Semantic correspondence is determined using the cosine similarity between the vector representations of the reference answer and the student’s response. An answer completeness indicator that accounts for the degree of coverage and the weights of key concepts is formalized. An integral indicator combining semantic correspondence and content completeness is proposed. The scientific result lies in the further development of methods for the automated analysis of open-ended responses by using identified content gaps to select an adaptive recommendation. The practical significance lies in the possibility of applying the method in distance learning systems to provide targeted feedback.
- 25 June 2026
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 66,p. 4-9
Open AccessMethod for generating adaptive learning recommendations based on semantic analysis of students’ responses
This study is devoted to developing a method for improving the efficiency of the educational process through the automated analysis of students’ responses and the generation of personalized recommendations. The object of the study is the process of generating adaptive learning recommendations based on the automated analysis of students’ open-ended textual responses. The subject of the study comprises models and methods of neural-network-based semantic analysis that enable the determination of the semantic correspondence between a student’s response and a reference answer, the completeness of key concept coverage, and the presence of content gaps. The aim of the study is to provide a theoretical justification and formalization of a method for generating adaptive learning recommendations based on the semantic analysis of students’ responses. A Siamese transformer neural network based on Sentence-BERT is proposed for generating vector representations of texts, enabling the consideration of context, synonymous constructions, and paraphrasing. Semantic correspondence is determined using the cosine similarity between the vector representations of the reference answer and the student’s response. An answer completeness indicator that accounts for the degree of coverage and the weights of key concepts is formalized. An integral indicator combining semantic correspondence and content completeness is proposed. The scientific result lies in the further development of methods for the automated analysis of open-ended responses by using identified content gaps to select an adaptive recommendation. The practical significance lies in the possibility of applying the method in distance learning systems to provide targeted feedback.
- Open Access
- 26 March 2026
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 65
Conceptual foundations for creating digital twins of damaged infrastructure objects
The article explores the issues of integrating digital twins, AI-agent systems, and multimodal human-computer interaction in the tasks of monitoring and dynamic verification of the state of damaged infrastructure objects. The relevance of the work is driven by the increasing complexity of modern infrastructure systems, the need for rapid response to damage, the fragmentation of information flows in crisis conditions, and heightened requirements for the reliability, transparency, and validity of management decisions. Traditional monitoring approaches do not provide a sufficient level of integration of heterogeneous data sources and adaptability to environmental changes, which necessitates the use of intelligent cyber-physical solutions. The study systematizes modern scientific approaches to building digital twins for civil infrastructure, analyzes the role of AI-agent mechanisms in coordinating analytical procedures, and generalizes the conceptual foundations and development trends of multimodal interaction in intelligent information systems. Particular attention is paid to the formation of the architectural structure for creating a digital twin of a damaged object based on the integration of sensory, visual, geospatial, and expert data, followed by their synchronization, normalization, structural binding, and dynamic verification procedure. The result of the study is the formation of the conceptual foundations of an integrated system where the digital twin acts as the core of the object's information model, the AI-agent system provides adaptive analytical support, coordination of processing and data reconciliation, and multimodal interaction creates a human-oriented mechanism for including an expert in the decision-making system loop. Key problematic aspects related to the reliability of digital models under uncertainty, multi-agent coordination in distributed environments, explainability of analytical results, synchronization of interaction modalities, and standardization of integration architectures are identified. The generalization of the conducted research suggests that the integration of digital twins, AI-agent technologies, and multimodal interaction forms the conceptual prerequisites for increasing the resilience, scalability, adaptability, and efficiency of managing the development of complex infrastructure objects in a dynamic and uncertain environment.
- 26 March 2026
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 65
Open AccessConceptual foundations for creating digital twins of damaged infrastructure objects
The article explores the issues of integrating digital twins, AI-agent systems, and multimodal human-computer interaction in the tasks of monitoring and dynamic verification of the state of damaged infrastructure objects. The relevance of the work is driven by the increasing complexity of modern infrastructure systems, the need for rapid response to damage, the fragmentation of information flows in crisis conditions, and heightened requirements for the reliability, transparency, and validity of management decisions. Traditional monitoring approaches do not provide a sufficient level of integration of heterogeneous data sources and adaptability to environmental changes, which necessitates the use of intelligent cyber-physical solutions. The study systematizes modern scientific approaches to building digital twins for civil infrastructure, analyzes the role of AI-agent mechanisms in coordinating analytical procedures, and generalizes the conceptual foundations and development trends of multimodal interaction in intelligent information systems. Particular attention is paid to the formation of the architectural structure for creating a digital twin of a damaged object based on the integration of sensory, visual, geospatial, and expert data, followed by their synchronization, normalization, structural binding, and dynamic verification procedure. The result of the study is the formation of the conceptual foundations of an integrated system where the digital twin acts as the core of the object's information model, the AI-agent system provides adaptive analytical support, coordination of processing and data reconciliation, and multimodal interaction creates a human-oriented mechanism for including an expert in the decision-making system loop. Key problematic aspects related to the reliability of digital models under uncertainty, multi-agent coordination in distributed environments, explainability of analytical results, synchronization of interaction modalities, and standardization of integration architectures are identified. The generalization of the conducted research suggests that the integration of digital twins, AI-agent technologies, and multimodal interaction forms the conceptual prerequisites for increasing the resilience, scalability, adaptability, and efficiency of managing the development of complex infrastructure objects in a dynamic and uncertain environment.
- Open Access
- 28 March 2025
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 61,p. 180-186
Models and methods of artificial intelligence in the process of performing building-technical expertise
The object of this research is the process of forming a conclusion building-technical expertise by a support system for the restoration of real estate objects. The subject of the study includes artificial intelligence models and methods capable of solving the task of forming an expert conclusion regarding the technical condition category of building structures and objects as a whole. The aim of this work is to justify the choice of a model for assessing the technical condition of real estate objects in building-technical expertise based on an analysis of artificial intelligence models and methods that can handle fuzzy classification tasks. To evaluate the technical condition of building structures and objects, the use of gradient-boosted decision trees is proposed. This method corrects errors from previous iterations and considers the magnitude of different types of errors. It has been demonstrated that the iterative learning mechanism allows experts in building-technical expertise to refine or supplement the data on which conclusions are based. Adjustments to the conclusions of gradient-boosted decision tree ensembles can be made by experts in accordance with the regulatory framework. The input and output data of the model have been formalized, taking into account such an anthropogenic factor as the impact of weaponry. Five main structural elements have been identified, for each of which it is advisable to train decision tree ensembles. A loss function has been introduced that allows special attention to be given to the critical states of buildings and structures, where the risk of error may lead to total unsuitability or functional failure of structures or their elements. Based on the analysis of a series of studies, the choice of multi-agent theory has been justified as a subject for further research to ensure the scalability and flexibility of the support system for the restoration of real estate objects.
- 28 March 2025
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 61,p. 180-186
Open AccessModels and methods of artificial intelligence in the process of performing building-technical expertise
The object of this research is the process of forming a conclusion building-technical expertise by a support system for the restoration of real estate objects. The subject of the study includes artificial intelligence models and methods capable of solving the task of forming an expert conclusion regarding the technical condition category of building structures and objects as a whole. The aim of this work is to justify the choice of a model for assessing the technical condition of real estate objects in building-technical expertise based on an analysis of artificial intelligence models and methods that can handle fuzzy classification tasks. To evaluate the technical condition of building structures and objects, the use of gradient-boosted decision trees is proposed. This method corrects errors from previous iterations and considers the magnitude of different types of errors. It has been demonstrated that the iterative learning mechanism allows experts in building-technical expertise to refine or supplement the data on which conclusions are based. Adjustments to the conclusions of gradient-boosted decision tree ensembles can be made by experts in accordance with the regulatory framework. The input and output data of the model have been formalized, taking into account such an anthropogenic factor as the impact of weaponry. Five main structural elements have been identified, for each of which it is advisable to train decision tree ensembles. A loss function has been introduced that allows special attention to be given to the critical states of buildings and structures, where the risk of error may lead to total unsuitability or functional failure of structures or their elements. Based on the analysis of a series of studies, the choice of multi-agent theory has been justified as a subject for further research to ensure the scalability and flexibility of the support system for the restoration of real estate objects.
- Open Access
- 29 November 2024
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 60,p. 136-145
Support system real estate reconstruction process: processing and storage of data
This article is a continuation of the work aimed at developing an information and communication system to support the process of restoring real estate objects that have been damaged and destroyed as a result of military operations conducted by the russian federation with the support of the republic of belarus on the territory of Ukraine since February 24, 2022. A holistic view of the highly loaded distributed architecture of a scalable and fault-tolerant system for supporting the process of restoring real estate objects, designed to solve many problems of the construction industry of Ukraine, has been formed. A universal model of the data processing and storage module is proposed. This module is designed for the rapid processing, analysis and storage of a large amount of information presented in various formats, taking into account the need to process fuzzy textual information from the media and eyewitnesses, and visual data obtained from geographic information systems, satellite surveillance systems and unmanned aerial vehicles. The article shows a diagram of the sequence of interaction between services and the importance of ontology in solving the problems of overload and avoiding conflicts in understanding data coming from different sources. Based on the analysis of a number of studies in the form of optimization ideas, recommendations for potential improvement of the system ontology formation are provided. At the same time, the possibility of using the module in the process of ontology formation and ways to improve ontology-managed systems are shown. The practical significance of the work is to ensure continuous monitoring and timely updating of data on real estate objects, which is important for maintaining the relevance of information when performing restoration work.
- 29 November 2024
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 60,p. 136-145
Open AccessSupport system real estate reconstruction process: processing and storage of data
This article is a continuation of the work aimed at developing an information and communication system to support the process of restoring real estate objects that have been damaged and destroyed as a result of military operations conducted by the russian federation with the support of the republic of belarus on the territory of Ukraine since February 24, 2022. A holistic view of the highly loaded distributed architecture of a scalable and fault-tolerant system for supporting the process of restoring real estate objects, designed to solve many problems of the construction industry of Ukraine, has been formed. A universal model of the data processing and storage module is proposed. This module is designed for the rapid processing, analysis and storage of a large amount of information presented in various formats, taking into account the need to process fuzzy textual information from the media and eyewitnesses, and visual data obtained from geographic information systems, satellite surveillance systems and unmanned aerial vehicles. The article shows a diagram of the sequence of interaction between services and the importance of ontology in solving the problems of overload and avoiding conflicts in understanding data coming from different sources. Based on the analysis of a number of studies in the form of optimization ideas, recommendations for potential improvement of the system ontology formation are provided. At the same time, the possibility of using the module in the process of ontology formation and ways to improve ontology-managed systems are shown. The practical significance of the work is to ensure continuous monitoring and timely updating of data on real estate objects, which is important for maintaining the relevance of information when performing restoration work.
- Open Access
- 24 November 2023
- Published by Kyiv National University of Construction and Architecture in WAYS TO IMPROVE CONSTRUCTION EFFICIENCY
- Vol. 3 (52) , 207-216
Development paths of ontology-driven support systems for building-technical expertise
The object of the research is the process of assessing the technical condition of objects within the scope of building-technical expertise. The subject of the study is the ontological approach to knowledge formalization and machine learning methods capable of solving the problem of evaluating the technical condition of damaged or destroyed structures. It is demonstrated that the increasing scale of physical destruction, the absence of a structured mechanism for processing expert information, and the need for coordination among different institutions highlight the necessity of developing a formalized and normatively consistent decision support system for experts. The paper explores the applicability of the ontological approach to the formal representation of knowledge in the field of construction and technical assessment. The structure of the ontology is outlined, along with its components and the stages of its development. A taxonomic model has been constructed, defining classes of structures, structural elements, types of damage, sets of influencing factors, and sources of the regulatory framework. The integration of machine learning methods as a tool for inductive data generalization within the ontological framework is substantiated. Examples of relevant algorithms are provided, including gradient boosting of decision trees, random forest, support vector machines, and the k-nearest neighbors method. It is emphasized that the system must ensure transparency, explainability, and compliance with regulatory requirements, while leaving the final decision to the domain expert. The results of this study can be used to develop a decision support system for assessing the technical condition of structures and formulating recommendations for their restoration.
- 24 November 2023
- Published by Kyiv National University of Construction and Architecture in WAYS TO IMPROVE CONSTRUCTION EFFICIENCY
- Vol. 3 (52) , 207-216
Open AccessDevelopment paths of ontology-driven support systems for building-technical expertise
The object of the research is the process of assessing the technical condition of objects within the scope of building-technical expertise. The subject of the study is the ontological approach to knowledge formalization and machine learning methods capable of solving the problem of evaluating the technical condition of damaged or destroyed structures. It is demonstrated that the increasing scale of physical destruction, the absence of a structured mechanism for processing expert information, and the need for coordination among different institutions highlight the necessity of developing a formalized and normatively consistent decision support system for experts. The paper explores the applicability of the ontological approach to the formal representation of knowledge in the field of construction and technical assessment. The structure of the ontology is outlined, along with its components and the stages of its development. A taxonomic model has been constructed, defining classes of structures, structural elements, types of damage, sets of influencing factors, and sources of the regulatory framework. The integration of machine learning methods as a tool for inductive data generalization within the ontological framework is substantiated. Examples of relevant algorithms are provided, including gradient boosting of decision trees, random forest, support vector machines, and the k-nearest neighbors method. It is emphasized that the system must ensure transparency, explainability, and compliance with regulatory requirements, while leaving the final decision to the domain expert. The results of this study can be used to develop a decision support system for assessing the technical condition of structures and formulating recommendations for their restoration.
- 4 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 532-538
Modeling the Process of Assessing the Technical Condition of Damaged Real Estate Objects
More than 160 thousand properties have already been registered on the territory of Ukraine as destroyed and damaged real estate objects as a result of hostilities. At present, houses, buildings and infrastructure objects in almost all regions of the country continue to experience damage and destruction. The reconstruction of so many objects that have suffered destruction and damage of different levels requires the creation of a publicly available, high-load distributed information and communication system to support decisions that will be made in the process of rebuilding of the country. The speed and reliability of such system depends on the architecture of its database. The data model for cataloging damaged buildings and structures, which will be used in MongoDB in the future, is given. The conclusion about the practicability of using MongoDB is made on the basis of an analysis of modern non-relational databases, taking into account the functional purposes of the Infocommunication System. The logical structure of the database and its main parts are defined and described. The scientific novelty of this study lies in the formation of fuzzy production models that adequately reproduce the associative thinking of experts when assessing the technical condition of objects that were affected by impact of hostilities that took place nearby.
- 4 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 532-538
Modeling the Process of Assessing the Technical Condition of Damaged Real Estate Objects
More than 160 thousand properties have already been registered on the territory of Ukraine as destroyed and damaged real estate objects as a result of hostilities. At present, houses, buildings and infrastructure objects in almost all regions of the country continue to experience damage and destruction. The reconstruction of so many objects that have suffered destruction and damage of different levels requires the creation of a publicly available, high-load distributed information and communication system to support decisions that will be made in the process of rebuilding of the country. The speed and reliability of such system depends on the architecture of its database. The data model for cataloging damaged buildings and structures, which will be used in MongoDB in the future, is given. The conclusion about the practicability of using MongoDB is made on the basis of an analysis of modern non-relational databases, taking into account the functional purposes of the Infocommunication System. The logical structure of the database and its main parts are defined and described. The scientific novelty of this study lies in the formation of fuzzy production models that adequately reproduce the associative thinking of experts when assessing the technical condition of objects that were affected by impact of hostilities that took place nearby.