Andrii Poliakov

Andrii Poliakov

Auto-generated profile, based on publicly available information

Andrii Poliakov

Simon Kuznets Kharkiv National University of Economics; Kharkiv National University of Radio Electronics

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 December 2024 in Eastern-European Journal of Enterprise Technologies
Verified
Journal Article
Open access
PDF file
HKHennadii Khudov
Hennadii Khudov
OMOleksandr Makoveichuk
Oleksandr Makoveichuk
OKOleksandr Kostyria
Oleksandr Kostyria
IBIhor Butko
Ihor Butko
APAndrii Poliakov
Andrii Poliakov
RKRostyslav Khudov
Rostyslav Khudov
The object of this study is the process of determining the coordinates of unmanned aerial vehicles. The study hypothesis assumed that the use of a network of portable spectrum analyzers could make it possible to detect the signals of the on-board systems of an unmanned aerial vehicle and reduce the mean square error in determining its coordinates. A method for determining the coordinates of an unmanned aerial vehicle using a network of portable spectrum analyzers has been improved, which, unlike known ones, allows for the following: – using signals of on-board equipment of an unmanned aerial vehicle; – using a network of portable spectrum analyzers; – application of both the triangulation and the difference-ranging method for determining the coordinates of an unmanned aerial vehicle by a network of portable spectrum analyzers; – carrying out spectral analysis of the signals of the on-board systems of the unmanned aerial vehicle (carried out additionally if necessary). Experimental studies have shown the capabilities of a portable spectrum analyzer to receive signals and display their spectra and spectrograms. The accuracy in determining the coordinates of an unmanned aerial vehicle by a network of portable spectrum analyzers was evaluated. It has been established: – the use of a network of portable spectrum analyzers significantly reduces the root mean square error in measuring the coordinates of an unmanned aerial vehicle by approximately 50 % compared to the error of one portable spectrum analyzer; – as the distance from the network elements of portable spectrum analyzers increases, the mean square error increases. – the use of a network of portable spectrum analyzers reduces the root mean square error in determining the coordinates of an unmanned aerial vehicle by an average of 2.29–2.62 times compared to the radar P-19MA, depending on the range

28 February 2025 in Eastern-European Journal of Enterprise Technologies
Verified
Journal Article
Open access
PDF file
HKHennadii Khudov
Hennadii Khudov
AHAndrii Hryzo
Andrii Hryzo
OOOleksandr Oleksenko
Oleksandr Oleksenko
IRIurii Repilo
Iurii Repilo
BLBohdan Lisohorskyi
Bohdan Lisohorskyi
SSSerhii Sukonko
Serhii Sukonko
The object of this study is the process of determining the coordinates of an air object. The main hypothesis of the study assumed that the use of a network of two Software-Defined Radio (SDR) receivers would make it possible to determine the coordinates of an air object. The determined coordinates could be used as preliminary target designation for the radar. A method for determining the coordinates of an air object by a network of two SDR receivers has been improved, which, unlike the known ones, involves: – using signals from the airborne systems of the air object; – using SDR receivers as network elements; – using the triangulation method for determining the coordinates of an air object. The accuracy of determining the coordinates of an air object by a network of two SDR receivers has been assessed. It was found that: – the accuracy of measuring the coordinates of an air object decreases sharply as the polar angle of observation from the middle of the base approaches 0 or π; – the smallest coordinate measurement error can be ensured when the air object is located on the traverse to the middle of the base and when the distance to the air object is close to the base size; – with small bases, the non-uniformity of the dependence of the coordinate determination error on the position of the air object relative to the SDR receivers is more pronounced than with large ones; – at a long range, the error values for small bases grow rapidly, which is primarily due to the small angle of intersection of the bearing lines; – it is advisable to place SDR receivers at a sufficiently large distance from each other (recommended value: (1–3) times the distance to the air object); – the errors of measuring the coordinates of the air object have a value of (250–350) m in a sufficiently wide range of directions; – with a decrease in the base size, the errors grow rapidly (reaching a value of more than 1000 m) when the observation angle deviates from the 90° direction

18 March 2020 in Системи обробки інформації
Verified
Journal Article
Open access
PDF file
АГА.о. Гапон
А.о. Гапон
ВФВ.м. Федорченко
В.м. Федорченко
АПА.о. Поляков
А.о. Поляков
Гарантування безпеки програмного продукту з відкритим вихідним кодом є актуальною проблемою, бо навіть у проектах з закритим вихідним кодом можуть бути присутні open source бібліотеки, що робить можливим появу вразливості у них. Серед методів, що використовують для виявлення вразливостей, варто виділити моделювання загроз, бо цей метод дозволяє вже на ранніх етапах розробки програмного коду прийняти заходи, що знизять витрати на ліквідацію вразливостей та спростять їх усунення і зміни до архітектури додатку. Підбір відповідного підходу при побудові моделі загроз залежить від специфіки проекту, ресурсів, а також кваліфікації адміністраторів.

30 June 2025 in Innovative technologies and scientific solutions for industries
Verified
Journal Article
Open access
PDF file
VSViacheslav Shkurko
Viacheslav Shkurko
APAndrii Poliakov
Andrii Poliakov
The article addresses the problem of organizing software and neural network algorithms for natural language processing. Given the rapid growth of information flows and the limitations of computational resources, optimizing methods for processing short, unstructured messages and complex structured documents has become especially urgent. This study aims to develop a comprehensive method for organizing computer-aided analysis of text data, ensuring a balance between the accuracy of results and the efficiency of computational resource use. The research systematically examines modern approaches to tokenization, clustering, semantic-relevant search, and deep learning architectures, with particular attention to their adaptability under resource-constrained conditions. A multilevel methodology is suggested, combining the preliminary classification of text arrays, semantic clustering, and the use of a Bidirectional LSTM neural network model. Practical implementation of the method was tested through an automated text analysis application, demonstrating stable reduction in the loss function and acceptable resource consumption. The ability to adapt to different types of text data, reduced resource consumption while maintaining high-quality analysis, and suitability for deployment in environments with low computing capacity should be considered the main advantages of the developed approach. The scientific novelty of the article is substantiated by the integration of semantic-relevant clustering and lightweight deep learning techniques into a single optimized framework. The practical value of the study is the possibility of applying the proposed methodology in real-world information systems where limited hardware capabilities require efficient and adaptive text processing solutions.

30 October 2026 in Eastern-European Journal of Enterprise Technologies
Verified
Journal Article
Open access
HKHennadii Khudov
Hennadii Khudov
VKVladyslav Khudov
Vladyslav Khudov
OMOleksandr Makoveichuk
Oleksandr Makoveichuk
IKIrina Khizhnyak
Irina Khizhnyak
AKAnastasiia Kirnos
Anastasiia Kirnos
AKAndrii Kolodka
Andrii Kolodka
The process that segments an image acquired from an onboard optoelectronic surveillance system has been considered in this study. The task addressed relates to the fact that established segmentation methods based on ant algorithms do not take into account the influence of pixel local neighborhood topology on the set of permissible transitions for agents (ants). This leads to discontinuities in the contour of small objects of interest and to an increase in segmentation errors of the first and second kind. This work has improved a method for segmenting an optoelectronic image from an onboard optoelectronic surveillance system based on the maxmin ant algorithm, according to which, in contrast to established ones: – when constructing the graph of permissible transitions, a pixel's local neighborhood topology (von Neumann topology and Moore topology) is taken into account; – the construction of agent (ants) paths and the update of the pheromone are carried out taking into account the specified topology; – the pheromone is added to the edges that belong to the best paths of agents (ants). An experimental study has been conducted on the segmentation of optoelectronic images using the method based on the maxmin ant algorithm with different topologies of the local neighborhood of the pixel. It was found that when segmenting optoelectronic images using the Moore topology, the object of interest (Lancet-type unmanned aerial vehicle) was selected in all five runs of the method. When segmenting optoelectronic images using the von Neumann topology, the object of interest (Lancet-type unmanned aerial vehicle) was selected in only two out of five runs of the method. For the segmented image, the segmentation error of the second kind was reduced from (44.36 ± 36.67) % (method using the von Neumann topology) to 0% (method using the Moore topology). Compared with the Otsu segmentation method, the segmentation errors of the first and second kinds were reduced by 1.51% (method using the von Neumann topology) and 4.92% (method using the Moore topology). Of practical importance is the fact that the improved method makes it possible to reduce segmentation errors of objects of interest without using standards and prior training. The scope of application includes airborne optoelectronic surveillance and air situation monitoring systems

31 March 2025 in Innovative technologies and scientific solutions for industries
Verified
Journal Article
Open access
PDF file
DHDenys Herasymuk
Denys Herasymuk
APAndrii Poliakov
Andrii Poliakov
VFVolodymy Fedorchenko
Volodymy Fedorchenko
The subject of the study in the article is the process of machine learning model selection performed by data scientists to build models in critical areas. The purpose of the work: 1) create a software library for measuring the accuracy, stability, and fairness of models; 2) conduct experiments to identify trade-offs between fairness, stability, and accuracy; 3) propose a responsible model selection process to improve the safety of using machine learning models. The article provides for the following tasks: to measure the fairness and stability of machine learning models and to investigate their relationship with the accuracy of models. The following methods are introduced: empirical evaluation, the theory of decomposition of model error into bias and variance, the theory of algorithmic fairness, and methods for quantitative assessment of uncertainty. Results achieved: 1) low predictive variability is proposed as a desirable property to ensure safety and equality of variability between different social groups as a new metric of fairness of machine learning models; 2) it is demonstrated how stability analysis helps specialists overcome the challenges of model multiplicity and choose reliable, stable and fair models; 3) an open-source software framework for community use is created that integrates stability measurements into model development processes. Conclusions. This work proposes a new paradigm of group fairness that combines the issues of correctness/quality and randomness/stability from the research agenda of responsible artificial intelligence. The application of the proposed approaches helps in the responsible selection of machine learning models under conditions of model multiplicity, demonstrating that, although there may be many models with comparable accuracy, there is only one (or a few) "best" model that is reliable, fair and stable, as it should be.