Oybek Usmonkulovich Mallaev

Dr. Oybek Usmonkulovich Mallaev

orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

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

Representative Publications

View All Publications
1 November 2019
Verified
Conference Paper
Closed access
ZXZayniddinov Xakimjon
Zayniddinov Xakimjon
Oybek Usmonkulovich MallaevOybek Usmonkulovich Mallaev
Oybek Usmonkulovich Mallaev

1 May 2020 in IOP Conference Series: Materials Science and Engineering
Verified
Conference Paper
Open access
HZH N Zaynidinov
H N Zaynidinov
Oybek Usmonkulovich MallaevOybek Usmonkulovich Mallaev
Oybek Usmonkulovich Mallaev
BAB B Anvarjonov
B B Anvarjonov
It is well-known that anyone who processes and analyzes image information needs a large amount of memory for storing images. For example, for storing a standard TV frame, if two of the three colors contain 2 bits for each of the 3 bits, the memory of at least 512×512 bytes is required. A passport image must be at least 64×64×64 bits (this requires more space than any other passport data). The problems associated with storage, search, and discovery are made more complex by processing and analyzing large numbers of images. The image contains a large amount of information, so it is natural for them to have excessive statistical, psychovisual and content information. Due to this redundancy, the process of digital image processing, image retrieval time, and storage capacity increase. The use of parallel technologies in digital processing of images plays an important role in solving the above problems. Therefore, the article presents an algorithm for calculating the process of extracting a human face from a color image using MPI technology.

30 November 2023
Verified
Conference Paper
Closed access
Oybek Usmonkulovich MallaevOybek Usmonkulovich Mallaev
Oybek Usmonkulovich Mallaev
JNJavohir Nurmurodov
Javohir Nurmurodov
SQSirojiddin Qobilov
Sirojiddin Qobilov
The use of neural networks for the processing of one- and two-dimensional data is examined in this research. The activating function being crucial to network performance, is frequently disregarded. This function is typically one of the non-adaptive functions. Adaptive sigmoid or ReLU functions are frequently employed but they have drawbacks as they alter the overall outcomes when data from a single, small area is adapted. Using flexible quadratic B-spline functions with free nodes is suggested in this study.

11 April 2023
Verified
Book Chapter
Closed access
PDF file
Oybek Usmonkulovich MallaevOybek Usmonkulovich Mallaev
Oybek Usmonkulovich Mallaev
BABunyodbek Azimov
Bunyodbek Azimov
KMKuchkarov Muslimjon
Kuchkarov Muslimjon
AKAhmadova Kamola
Ahmadova Kamola

26 February 2026 in Mathematics
Verified
Journal Article
Open access
PDF file
HKHalimjon Khujamatov
Halimjon Khujamatov
KIKobuljon Ismanov
Kobuljon Ismanov
Oybek Usmonkulovich MallaevOybek Usmonkulovich Mallaev
Oybek Usmonkulovich Mallaev
OSOtabek Sattarov
Otabek Sattarov
Cryptocurrency trading presents significant challenges due to extreme market volatility, rapid regime transitions, and non-stationary dynamics that render traditional trading strategies ineffective. Existing reinforcement learning approaches for cryptocurrency trading typically employ simplistic profit-based reward functions that fail to adequately capture risk management considerations, market microstructure costs, temporal dependencies, and regime-specific optimal behaviors. This limitation often results in strategies that perform well during favorable market conditions but suffer catastrophic losses during downturns. This paper introduces five novel reward functions grounded in economic utility theory, market microstructure, behavioral finance, adaptive risk management, and regime-conditional optimization. We systematically evaluate these reward functions across three reinforcement learning algorithms (Deep Q-Network, Proximal Policy Optimization, and Advantage Actor–Critic) and four distinct market regimes (bull, bear, high volatility, and recovery), using Bitcoin hourly data from 2018–2022. Our comprehensive experimental evaluation demonstrates that the Adaptive Risk Control reward function achieves exceptional performance, with a Sharpe ratio of 2.47, cumulative return of 26.4%, and maximum drawdown of only 16.8% during the predominantly bearish 2022 test period. Critically, regime-specific analysis reveals substantial performance heterogeneity: Adaptive Risk Control excels during high volatility (Sharpe ratio 3.21), while Temporal Coherence and Asymmetric Market-Conditional rewards dominate in trending and bear markets, respectively. These findings establish that sophisticated, theory-grounded reward engineering—rather than algorithmic innovations alone—constitutes the primary lever for improving RL trading systems, enabling positive risk-adjusted returns even during severe market downturns.

16 May 2025 in E3S Web of Conferences
Verified
Conference Paper
Open access
PDF file
Oybek Usmonkulovich MallaevOybek Usmonkulovich Mallaev
Oybek Usmonkulovich Mallaev
JGJamoliddin Gazatov
Jamoliddin Gazatov
JAJaloliddin Aliyev
Jaloliddin Aliyev
This article presents the development of a method, mechanism, module, and algorithm for the automatic analysis, real-time visualization, and intelligent resolution of citizen appeals using artificial intelligence (AI) and geo-mapping technologies. The article outlines the key directions and technologies involved in organizing interactive citizen appeal services and details the development process of an algorithm and software tool designed to create a geo-map of appeals requiring prompt attention. The proposed geo-map module offers essential functions such as automatic detection and analysis of appeals, identification and marking of problem locations on the map, prioritization of urgent issues, prediction and prevention of appeal surges, automation of state services, and rapid response facilitation. The module leverages AI for intelligent appeal analysis, employing Natural Language Processing to interpret appeal texts and classify them by category and priority-for example. Sentiment Analysis is used to assess the tone and urgency of appeals, assigning higher priority to urgent messages. Additionally, Machine Learning techniques predict the likelihood and geographic distribution of recurring problems based on historical electronic appeals-for instance, identifying increased gas supply complaints during winter months.