Anton Kovantsev
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Anton Kovantsev has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Alexander V. Boukhanovsky
ITMO University
5 shared publications - Kirill Zakharov
ITMO University
4 shared publications - Petr Chunaev
ITMO University
4 shared publications - Elizaveta Stavinova
ITMO University, Russian Federation
2 shared publications - A.G. Surikov1 shared publication
A.N. Kovatsev
1 shared publication
Latest Publications
- Open Access
- 10 October 2026
- Published by MDPI AG in Computer Sciences & Mathematics Forum
- Vol. 16 (1) , 20
Refinements of Reservoir Neural Networks for Chaotic Time-Series Forecasting
Forecasting chaotic and nonlinear time series in financial and economic domains remains an open challenge, as conventional statistical and machine learning methods struggle to capture their complex dynamics. We propose ESN-F, an Echo State Network enhanced with a Fourier transform layer and polynomial feature expansion, designed to improve forecasting accuracy for processes exhibiting chaotic, non-stationary, and non-normal behaviour. The model is evaluated on 1500 series drawn from the M4 Competition dataset and on a real-world Moscow real estate price index. On the M4 subset, ESN-F achieves a mean MAPE of 3.44% on predictable series and 5.26% on chaotic series, outperforming standard ESN, Light Gradient-Boosting Machine (LGBM), Prophet, and Singular Spectrum Analysis (SSA). On the real estate index, ESN-F reaches a cross-validated mean MAPE of 2.56%, compared with 3.19% for ESN and 7.18% for LGBM. In addition, we show that a set of dynamical features—including the Hurst exponent, Kolmogorov–Sinai entropy, correlation dimension, largest Lyapunov exponent, and dominant Fourier harmonics—can identify time series for which reservoir computing is most beneficial, enabling model selection prior to training. These results confirm that the proposed Fourier-augmented reservoir architecture provides a practical and computationally efficient approach to chaotic time-series forecasting.
- 10 October 2026
- Published by MDPI AG in Computer Sciences & Mathematics Forum
- Vol. 16 (1) , 20
Open AccessRefinements of Reservoir Neural Networks for Chaotic Time-Series Forecasting
Forecasting chaotic and nonlinear time series in financial and economic domains remains an open challenge, as conventional statistical and machine learning methods struggle to capture their complex dynamics. We propose ESN-F, an Echo State Network enhanced with a Fourier transform layer and polynomial feature expansion, designed to improve forecasting accuracy for processes exhibiting chaotic, non-stationary, and non-normal behaviour. The model is evaluated on 1500 series drawn from the M4 Competition dataset and on a real-world Moscow real estate price index. On the M4 subset, ESN-F achieves a mean MAPE of 3.44% on predictable series and 5.26% on chaotic series, outperforming standard ESN, Light Gradient-Boosting Machine (LGBM), Prophet, and Singular Spectrum Analysis (SSA). On the real estate index, ESN-F reaches a cross-validated mean MAPE of 2.56%, compared with 3.19% for ESN and 7.18% for LGBM. In addition, we show that a set of dynamical features—including the Hurst exponent, Kolmogorov–Sinai entropy, correlation dimension, largest Lyapunov exponent, and dominant Fourier harmonics—can identify time series for which reservoir computing is most beneficial, enabling model selection prior to training. These results confirm that the proposed Fourier-augmented reservoir architecture provides a practical and computationally efficient approach to chaotic time-series forecasting.
- 12 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 303-312
Realistic Controlled Time Series Generation with Variational Autoencoder
The controlled generation of time series facilitates the creation of scenarios for testing systems and models in unconventional situations, which also useful for enhancing the models generalizability. In this article, we presented an approach for realistic controlled time series generation based on a variational autoencoder, so-called ControlVAE. We modified the decoder by incorporating control elements. The adjustment is implemented using exponential smoothing and masking techniques, which enable the generation of realistic time series without abrupt discontinuities. Our model possesses the capability to handle three distinct scenarios that cover the most tasks: trend control, start and end point control, and value control at predetermined time. An experimental study conducted demonstrates that our approach accurately reproduces the patterns of the original time series, as well as the precise replication of scenarios while preserving the authenticity of synthetic time series.
- 12 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 303-312
Realistic Controlled Time Series Generation with Variational Autoencoder
The controlled generation of time series facilitates the creation of scenarios for testing systems and models in unconventional situations, which also useful for enhancing the models generalizability. In this article, we presented an approach for realistic controlled time series generation based on a variational autoencoder, so-called ControlVAE. We modified the decoder by incorporating control elements. The adjustment is implemented using exponential smoothing and masking techniques, which enable the generation of realistic time series without abrupt discontinuities. Our model possesses the capability to handle three distinct scenarios that cover the most tasks: trend control, start and end point control, and value control at predetermined time. An experimental study conducted demonstrates that our approach accurately reproduces the patterns of the original time series, as well as the precise replication of scenarios while preserving the authenticity of synthetic time series.
- 12 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1794-1801
Optimisation of a Socio-Economic System for Increasing Residents' Satisfaction Level
It is obvious that the problem of client migration in various businesses like banks, mobile operators and so on; the problem of user preferences' shift from one social network or computer game to another; or the problem of people's migration between cities and countries all have their roots in satisfaction and happiness provided by services or life conditions. The matching of these to human values, necessities, and identities makes a certain business or settlement successful and attractive. This task looks like an optimisation one. Optimisation tasks can be easily solved with a good deal of well-known algorithms. The only trouble in our case is the objective function, whose optimal value should be found. The ideal goal for optimisation should be satisfying clients' or residents' needs in various aspects of their life according to their customs, habits, values and identities, which is hard to measure and digitalise, like many things in the human world. In our research, we attempt to solve this task by means of cluster analysis for features discovery and determine the attractive business strategy or directions for development.
- 12 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1794-1801
Optimisation of a Socio-Economic System for Increasing Residents' Satisfaction Level
It is obvious that the problem of client migration in various businesses like banks, mobile operators and so on; the problem of user preferences' shift from one social network or computer game to another; or the problem of people's migration between cities and countries all have their roots in satisfaction and happiness provided by services or life conditions. The matching of these to human values, necessities, and identities makes a certain business or settlement successful and attractive. This task looks like an optimisation one. Optimisation tasks can be easily solved with a good deal of well-known algorithms. The only trouble in our case is the objective function, whose optimal value should be found. The ideal goal for optimisation should be satisfying clients' or residents' needs in various aspects of their life according to their customs, habits, values and identities, which is hard to measure and digitalise, like many things in the human world. In our research, we attempt to solve this task by means of cluster analysis for features discovery and determine the attractive business strategy or directions for development.
- 30 June 2025
- Published by Springer Nature
- p. 238-252
A Dynamic Model of Customers Behavior: Integrating Econophysics and Physics-Informed Neural Networks
- 30 June 2025
- Published by Springer Nature
- p. 238-252
A Dynamic Model of Customers Behavior: Integrating Econophysics and Physics-Informed Neural Networks
- Open Access
- 7 March 2025
- Published by MDPI AG in Smart Cities
- Vol. 8 (2) , 42
Coupling of Lagrangian Mechanics and Physics-Informed Neural Networks for the Identification of Migration Dynamics
An essential aspect of any government in a smart city is to examine the issues of internal and external migration. Migration is a complex phenomenon. In order to effectively manage it, it is not only necessary to be able to accurately predict migration patterns but also to understand which factors influence these patterns. Current approaches to the development of migration models rely on macroeconomic indicators without considering the specificities of intraregional interactions among individuals. In this paper, we propose a method for determining the dynamics of migration balance based on Lagrangian mechanics. We derive and interpret the potential energy of a migration network by introducing specific functions that determine migration patterns. The solution of the migration equations and selection of parameters, as well as external forces, are achieved through the use of physics-informed neural networks. We also use external factors to explain the non-homogeneity in the dynamic equation through the use of a regression model. We analyze settlement priorities using transfer operator theory and invariant density. The findings obtained enable the assessment of migration flows and analysis of external migration factors.
- 7 March 2025
- Published by MDPI AG in Smart Cities
- Vol. 8 (2) , 42
Open AccessCoupling of Lagrangian Mechanics and Physics-Informed Neural Networks for the Identification of Migration Dynamics
An essential aspect of any government in a smart city is to examine the issues of internal and external migration. Migration is a complex phenomenon. In order to effectively manage it, it is not only necessary to be able to accurately predict migration patterns but also to understand which factors influence these patterns. Current approaches to the development of migration models rely on macroeconomic indicators without considering the specificities of intraregional interactions among individuals. In this paper, we propose a method for determining the dynamics of migration balance based on Lagrangian mechanics. We derive and interpret the potential energy of a migration network by introducing specific functions that determine migration patterns. The solution of the migration equations and selection of parameters, as well as external forces, are achieved through the use of physics-informed neural networks. We also use external factors to explain the non-homogeneity in the dynamic equation through the use of a regression model. We analyze settlement priorities using transfer operator theory and invariant density. The findings obtained enable the assessment of migration flows and analysis of external migration factors.
- Open Access
- 3 September 2024
- Published by MDPI AG in Smart Cities
- Vol. 7 (5) , 2495-2513
Forecasting Population Migration in Small Settlements Using Generative Models under Conditions of Data Scarcity
Today, the problem of predicting population migration is essential in the concept of smart cities for the proper development planning of certain regions of the country, as well as their financing and landscaping. In dealing with population migration in small settlements whose population is below 100,000, data collection is challenging. In countries where data collection is not well developed, most of the available data in open access are presented as part of textual reports issued by authorities in municipal districts. Therefore, the creation of a more or less adequate dataset requires significant efforts, and despite these efforts, the outcome is far from ideal. However, for large cities, there are typically aggregated databases maintained by authorities. We used them to find out what factors had an impact on the number of people who arrived or departed the city. Then, we reviewed several dozens of documents to mine the data of small settlements. These data were not sufficient to solve machine learning tasks, but they were used as the basis for creating a synthetic sample for model fitting. We found that a combination of two models, each trained on synthetic data, performed better. A binary classifier predicted the migration direction and a regressor estimateed the number of migrants. Lastly, the model fitted with synthetics was applied to the other set of real data, and we obtained good results, which are presented in this paper.
- 3 September 2024
- Published by MDPI AG in Smart Cities
- Vol. 7 (5) , 2495-2513
Open AccessForecasting Population Migration in Small Settlements Using Generative Models under Conditions of Data Scarcity
Today, the problem of predicting population migration is essential in the concept of smart cities for the proper development planning of certain regions of the country, as well as their financing and landscaping. In dealing with population migration in small settlements whose population is below 100,000, data collection is challenging. In countries where data collection is not well developed, most of the available data in open access are presented as part of textual reports issued by authorities in municipal districts. Therefore, the creation of a more or less adequate dataset requires significant efforts, and despite these efforts, the outcome is far from ideal. However, for large cities, there are typically aggregated databases maintained by authorities. We used them to find out what factors had an impact on the number of people who arrived or departed the city. Then, we reviewed several dozens of documents to mine the data of small settlements. These data were not sufficient to solve machine learning tasks, but they were used as the basis for creating a synthetic sample for model fitting. We found that a combination of two models, each trained on synthetic data, performed better. A binary classifier predicted the migration direction and a regressor estimateed the number of migrants. Lastly, the model fitted with synthetics was applied to the other set of real data, and we obtained good results, which are presented in this paper.
- 29 February 2024
- Published by Springer Nature
- p. 80-90
Consumer Behaviour Timewise Dependencies Investigation by Means of Transition Graph
- 29 February 2024
- Published by Springer Nature
- p. 80-90
Consumer Behaviour Timewise Dependencies Investigation by Means of Transition Graph
- Open Access
Enhancing Predictability Assessment: An Overview and Analysis of Predictability Measures for Time Series and Network Links
Driven by the variety of available measures intended to estimate predictability of diverse objects such as time series and network links, this paper presents a comprehensive overview of the existing literature in this domain. Our overview delves into predictability from two distinct perspectives: the intrinsic predictability, which represents a data property independent of the chosen forecasting model and serves as the highest achievable forecasting quality level, and the realized predictability, which represents a chosen quality metric for a specific pair of data and model. The reviewed measures are used to assess predictability across different objects, starting from time series (univariate, multivariate, and categorical) to network links. Through experiments, we establish a noticeable relationship between measures of realized and intrinsic predictability in both generated and real-world time series data (with the correlation coefficient being statistically significant at a 5% significance level). The discovered correlation in this research holds significant value for tasks related to evaluating time series complexity and their potential to be accurately predicted.
Open AccessEnhancing Predictability Assessment: An Overview and Analysis of Predictability Measures for Time Series and Network Links
Driven by the variety of available measures intended to estimate predictability of diverse objects such as time series and network links, this paper presents a comprehensive overview of the existing literature in this domain. Our overview delves into predictability from two distinct perspectives: the intrinsic predictability, which represents a data property independent of the chosen forecasting model and serves as the highest achievable forecasting quality level, and the realized predictability, which represents a chosen quality metric for a specific pair of data and model. The reviewed measures are used to assess predictability across different objects, starting from time series (univariate, multivariate, and categorical) to network links. Through experiments, we establish a noticeable relationship between measures of realized and intrinsic predictability in both generated and real-world time series data (with the correlation coefficient being statistically significant at a 5% significance level). The discovered correlation in this research holds significant value for tasks related to evaluating time series complexity and their potential to be accurately predicted.
- Open Access
- 5 July 2023
- Published by MDPI AG in Engineering Proceedings
- Vol. 39 (1) , 61
Enhancement of Consumption Forecasting by Customers’ Behavioral Predictability Segregation
The easiest approach to customer activity forecasting involves using the whole available and applicable population of customers that a certain data set contains. The drawback of this simple technique is twofold: the set could be too big, and it could contain customers of very different peculiarities, which means that customers whose previous behavior is helpful for the forecast and whose one is not are mixed, and while the first performs a good-quality prediction, the second spoils it by adding noise. Hence, if we could choose the customers with good predictability and put aside the others “as a shepherd divideth his sheep from the goats” (Matthew 25:32), we would solve both problems: less data volume and less noise; the principle is like ancient “divide et impera”. In our research, we developed the method of customers separation by predictability and its dynamics with the help of LSTM models. Our research shows that (1) customer separation helps to improve the forecasting quality of the whole population due to the decomposition of all clients’ time series, and (2) environmental instability such as pandemics or military action can be leveled out with incremental models.
- 5 July 2023
- Published by MDPI AG in Engineering Proceedings
- Vol. 39 (1) , 61
Open AccessEnhancement of Consumption Forecasting by Customers’ Behavioral Predictability Segregation
The easiest approach to customer activity forecasting involves using the whole available and applicable population of customers that a certain data set contains. The drawback of this simple technique is twofold: the set could be too big, and it could contain customers of very different peculiarities, which means that customers whose previous behavior is helpful for the forecast and whose one is not are mixed, and while the first performs a good-quality prediction, the second spoils it by adding noise. Hence, if we could choose the customers with good predictability and put aside the others “as a shepherd divideth his sheep from the goats” (Matthew 25:32), we would solve both problems: less data volume and less noise; the principle is like ancient “divide et impera”. In our research, we developed the method of customers separation by predictability and its dynamics with the help of LSTM models. Our research shows that (1) customer separation helps to improve the forecasting quality of the whole population due to the decomposition of all clients’ time series, and (2) environmental instability such as pandemics or military action can be leveled out with incremental models.
- Open Access
- 1 February 2023
- Published by ITMO University in Scientific and Technical Journal of Information Technologies, Mechanics and Optics
- Vol. 23 (1) , 121
Value-based modeling of economic decision making in conditions of unsteady environment
The changing environment creates the conditions for changing people’s behavior which together can lead to crisis situations. In the case of making economic decisions, the emerging non-stationarity of the dynamics of different components of the system can represent an economic crisis. The possibility of universal human values consideration in modeling decision-making under conditions of changing environment has been studied. The internal processes of an agent prior to making a decision are reflected in the concept of Beliefs-Desires-Intentions-Actions-Reactions (BDIAR). The article reviews the regularities and existing approaches to modeling economic decisions, and proposes a new, author’s approach. The mechanisms of influence of stress on decision-making, factors of rationality limitation, and risk assessment in the context of behavioral economics are revealed. The well-known theories of the influence of values on the change in the structure of consumption in a crisis situation are considered. Ways of taking into account emotions in the architecture of the agent are shown. In the proposed model, values are considered as a social factor in decision making. Due to their subjectivity, they are presented mathematically as the basis for assessing environmental objects. The subjectivity of the objects of choice assessments is reflected in the functions of attractiveness of the objects, the functions of the agent’s state dynamics, and in the subjectivity of the decision influence on the satisfaction of the agent. A possible modification of the components of the agent model is shown, taking into account the influence of values on the dynamics of consumption. A method is proposed for taking into account values in the BDIAR architecture when modeling an agent’s decision-making where the levels of the architecture correspond to values, preference functions, and functions of the agent’s state dynamics. The pseudonymized transactional data on debit cards of the partner bank customers were analyzed separately for 2017–2019 and 2020. The subjectivity of the environment influence in a crisis situation on the dynamics of changes in values and needs for different consumer groups is demonstrated, taking into account the clustering of their behavior types. Differences in the dynamics of values at the individual level and the level of groups are shown; an increase in the priority of survival values and a different rate of return to the pre-crisis state for different behavioral groups are also shown. The results obtained are useful for developing methods for modeling economic behavior in a non-stationary external environment, in particular, in the case of crises.
- 1 February 2023
- Published by ITMO University in Scientific and Technical Journal of Information Technologies, Mechanics and Optics
- Vol. 23 (1) , 121
Open AccessValue-based modeling of economic decision making in conditions of unsteady environment
The changing environment creates the conditions for changing people’s behavior which together can lead to crisis situations. In the case of making economic decisions, the emerging non-stationarity of the dynamics of different components of the system can represent an economic crisis. The possibility of universal human values consideration in modeling decision-making under conditions of changing environment has been studied. The internal processes of an agent prior to making a decision are reflected in the concept of Beliefs-Desires-Intentions-Actions-Reactions (BDIAR). The article reviews the regularities and existing approaches to modeling economic decisions, and proposes a new, author’s approach. The mechanisms of influence of stress on decision-making, factors of rationality limitation, and risk assessment in the context of behavioral economics are revealed. The well-known theories of the influence of values on the change in the structure of consumption in a crisis situation are considered. Ways of taking into account emotions in the architecture of the agent are shown. In the proposed model, values are considered as a social factor in decision making. Due to their subjectivity, they are presented mathematically as the basis for assessing environmental objects. The subjectivity of the objects of choice assessments is reflected in the functions of attractiveness of the objects, the functions of the agent’s state dynamics, and in the subjectivity of the decision influence on the satisfaction of the agent. A possible modification of the components of the agent model is shown, taking into account the influence of values on the dynamics of consumption. A method is proposed for taking into account values in the BDIAR architecture when modeling an agent’s decision-making where the levels of the architecture correspond to values, preference functions, and functions of the agent’s state dynamics. The pseudonymized transactional data on debit cards of the partner bank customers were analyzed separately for 2017–2019 and 2020. The subjectivity of the environment influence in a crisis situation on the dynamics of changes in values and needs for different consumer groups is demonstrated, taking into account the clustering of their behavior types. Differences in the dynamics of values at the individual level and the level of groups are shown; an increase in the priority of survival values and a different rate of return to the pre-crisis state for different behavioral groups are also shown. The results obtained are useful for developing methods for modeling economic behavior in a non-stationary external environment, in particular, in the case of crises.