CENTRAL ASIAN JOURNAL OF MATHEMATICAL THEORY AND COMPUTER SCIENCES

Traffic Sign Recognition Using Convolutional Neural Networks with a Tkinter-Based Interactive Interface

2026-02-11T10:32:46+00:00

This project demonstrates a traffic sign recognition system that employs Convolutional Neural Networks (CNN) to categorise traffic signs from the dataset. The code has two main parts: training the CNN model and a graphical user interface (GUI) made with Tkinter that lets you interact with the model. The code prepares the dataset for the CNN model training phase by putting images and labels into numpy arrays. For model evaluation, the data is divided into training and testing sets. To effectively capture traffic sign features, the model architecture has several convolutional and pooling layers, followed by fully connected layers. The Adam optimiser and categorical cross-entropy loss are used to train the model. You can see how well the training is going by looking at the accuracy and loss curves. The code has a Tkinter-based GUI that lets users upload pictures of traffic signs so that they can be sorted. The GUI loads the CNN model that has already been trained and uses a pre-defined dictionary to map the predicted class to the description of the traffic sign. The system guesses what kind of traffic sign the user uploaded and shows its description on the GUI. This traffic sign recognition system has an easy-to-use interface for classifying traffic signs in real time. This can be helpful for learning, making roads safer, and analysing traffic sign data. The CNN model is accurate, and the Tkinter GUI is easy to use, so the system can be used on a variety of platforms.

A Comparison Between Bootstrap and Bayesian Methods In Estimating the Parameters of the Conditional Logistic Regression Model with A Practical Application

2026-02-10T04:26:57+00:00

Objects can arrive at a conclusion by two logical and inferential procedures: indistinguishable-to-average consecutive likelihood and traditional evaluation, the test would be satisfactory. This paper tests two kinds of estimation techniques: Bootstrap resampling and Bayesian inference, for estimating the coefficients in a Conditional Logistic Regression model applied to heart disease data. Classical estimates were obtained using the clogit (CL) function, and the Bootstrap estimates came from 100 resampled samples. Both methods give similar estimates, with the Bootstrap method supplying an additional advantage of quantifying variability. There was an attempt of Bayesian estimation but it failed as the program could not properly initialize in the present environment. This gives further support to the use of Bootstrap and Classical methods for practical analysis, and lays a foundation for Bayesian treatment in future development efforts.

Enhancing Organizational Efficiency through AI-‎Driven Information Systems Management

2026-01-24T00:21:47+00:00

هدف البحث: تحسين الأداء الإداري والتنظيمي للمؤسسات من خلال واجهة حوارية مصممة خصيصًا، حيث يُدخل المستخدم بياناته فتظهر النتائج، مع التركيز على العناصر الأكثر تأثيرًا في هذا الأداء، وتقديم رؤى مستقبلية.
أدوات البحث: تم استخدام أسلوب الغابة العشوائية ولغة البرمجة بايثون، ودُرست مجموعة من العوامل من مؤسسات ذات تخصصات مختلفة.
نتائج البحث: حقق أسلوب الغابة العشوائية دقة عالية بلغت 92%. كما كان لهذا النموذج المصمم أثرٌ كبير في رفع جودة العمل في المؤسسات من خلال تقديم رؤى مستقبلية وتحديد العوامل الأكثر تأثيرًا على جودة العمل، وأهمها الكفاءة الإنتاجية.

Sustainable Cloud-Based Security Architecture for Data Protection in E-Learning and Healthcare Systems

2026-02-16T22:02:21+00:00

Abstract: Cloud computing is one of the new innovations in digital transformation that have revolutionized various areas of life including education and healthcare. However, with the rising significance of cloud infrastructures, the level of security threat and privacy concern is on the rise with regard to sensitive data. Sustainable cloud encryption techniques will be discussed in this paper in the enhancement of data security in e-learning and health-related platform.

The hybrid encryption techniques, proxy re-encryption and block-chain-based systems, are analyzed with the assistance of systematic literature review. It offers a proposed security structure which is based on federated learning, with sustainable encryption to strengthen data protection. The findings have shown that sustainable encryption is a means to ensure confidentiality, integrity, as well as minimizing the cost of computing and environmental costs thereby ensuring the digital ecosystems are secure and environmental friendly.

Engineering Trustworthy Software with Large Language Models: A Hybrid Framework for Automated Testing, Repair, and Reliability Assurance

2026-02-21T08:27:36+00:00

Large Language Models (LLMs) have recently become powerful automation enablers in
software engineering due to their outstanding capabilities in code synthesis, automatic
programming and program fixing. Notwithstanding of all these advances, probabilistic nature of
LLMs raises substantial concerns about the software accuracy, reliability as well as trustworthiness
in long term, particularly when such models become deployable without an engineering
supervision or a systematic validation. We are missing an essential step in reliability-centered and
system-level engineering because existing research mainly investigates single LLM-assisted tasks
and frequently assume task-level performance capabilities. Through combining LLM-based
automation with traditional software adversarial challenging, end-to-end program repair and
reliability attestations, this work provides a potential hybridized framework to produce reliable
software using Large Language Models. To systematically control the proliferation of LLMs through
the software development life-cycle, the framework provides a deterministic authentication
pipeline, reliability-sensitive valuation metrics and feedback-based adaptation loops. We
conducted an extensive empirical evaluation, comparing the proposed framework against: (i)
baseline tools from existing work; (ii) traditional automated repair techniques, and (iii) unrestricted
plans based on LLM. Results indicate that the hybrid approach significantly enhances test coverage
and discovery of faults, yields a higher proportion of semantically correct patches with greatly
mitigated over fitting, and improves the maintainability and fault recurrence for better long term
software reliability. These results lead to question whether pure LLM does not indeed guarantees
a robust AI assisted software engineering. It's those hybrid models that you want - ones which meld
deterministic software engineering with a grain of management craft. We formulate and prove both
the dictionary based theoretical observation and its empirical counterpart in the context of Large
Language Models, in so far as is possible with robust software engineering for systems at that scale.

The Role and Advantages of Cloud Technologies in Information Systems

2026-02-22T03:32:12+00:00

In this paper we study the role of cloud tools in the field developing of modern information systems.ессе We analyze the advantages provided by cloud technologies in comparison with traditional ones and highlights their increasing relevance in terms digital transformation. Cloud computing is discussed as a new approach of providing resources (including computing resource, storage and software service) over the internet. The paper elaborates the core service models as IaaS, PaaS and SaaS and assesses their practical relevance for enterprises or institutions from different domains. I focus, in particular, on the technological underpinnings of cloud infrastructure such as virtualization, distributed computing and resource pooling. The report outlines the benefits of cloud solutions, including cost control, scalability, flexibility, availability and better security of user data. Cloud solutions facilitate dynamic provisioning of computing capabilities, and they implement a PAYG model to decrease CAPEX and to improve the OPEX efficiency. The study also discusses perceived blockers to the cloud - such as cyber security threats, data privacy and regulations. By means of descripto-analytical and comparison methods that lead one to understand that today, cloud technologies not only ameliorate IT infrastructure, but also play as a strategic devices for innovation and sustainable development. The results support the view that cloud computing is highly influential in enhancing organizational performance and maintaining long term competitiveness in a fast paced digital environment.

Neural Style Transfer: Integrating Deep Learning Techniques with Artistic and Cultural Expression

2026-02-23T20:20:26+00:00

Neural Style Transfer (NST) has recently emerged as a revolutionary trend at the intersection of profound literacy and cultural expression. This new idea is based on how people naturally create things. It lets you blend content from one image with the style of another, creating a new, visually appealing image. This book talks about the ideas, methods, and workings of NST. NST is a new way to make art that combines photos with rich content and the unique styles of famous painters, illustrators, and others. It does this by routing and manipulating point representations from pre-trained networks. The process optimises a total loss function that combines content and style losses. This lets images keep both the substance of the content source and the stylistic details of the reference image. This book goes into detail about the specialised parts of NST, showing how convolutional layers in deep neural networks capture the content and style of images. We talk about how to use loss functions and the iterative optimisation process to make beautiful compositions. We also examine how hyperparameters and loss weighting affect the transfer of content and style, enabling us to exert more precise control. This work shows the wide range of operations NST can perform, in addition to its specialised ones. NST has made progress in many areas, including graphic design, fine arts, and computer vision. They have done everything from reimagining photos as if painted by expressionist masters to creating new textures and designs. This publication gives useful examples and real-world use cases that show how NST could be used in the future. NST opens new ways of talking about culture and provides both artists and technologists with valuable tools. It can turn everyday images into works of art.

Identification of Fungi on Mobile Phone Surfaces and Evaluation of the Effectiveness of Antibiotics Against Them

2026-02-14T14:13:00+00:00

Mobile phones are a means by which many pathogens, including fungi, are transmitted. Therefore, this research examined the insulation and recognition of fungi isolated from mobile phones used by university professors and staff, as well as doctors, pharmacists, taxi drivers, and restaurant workers. A total of 80 samples were collected from 2024-12-1, to 2025-2-1. The study addressed the virulence factors of Aspergillus chevalieri, one of the fungi isolated from mobile phones, which include the capacity to adhere, the capacity to make blood disintegrate, and the synthesis of proteolytic enzymes, lipids, and proteases. Drug sensitivity of A. chevalieri to antifungals was tested. The study's findings demonstrated the existence of fungi upon mobile phones, as fungus Aspergillus spp was prevalent in all types of phones, and the species appeared in A. fumigatus 17 at a rate of (17%), as the highest rate was recorded, followed by the fungus A. niger at a rate of (13%), the fungus A. candidus at a rate of (12%), the fungus A. chevalieri and A. flavus at a rate of (11%), the fungus A. terreus eight isolates at a rate of (8%), seven isolates Penicillium spp at a rate of (7%), six isolates of Mucor spp, three isolates of Alternaria spp, two isolates of each fungus, Cladosporium spp and C. parapsilosis, and four isolates of each fungus Paecilomyces spp, Rhizopus spp. Molecular identification of A. chevalieri revealed that it contained a 500 bp DNA band. The results of the genetic tree analysis and registration on the global GenBank website also showed a similarity between our isolates and global isolates. A. chevalieri's virulence factors were examined, and the findings demonstrated that spores could cling to epithelial cells by 25%. Regarding the capacity to dissolve blood, favorable outcomes were shown, as it was an alpha-type hemolysis, and it could produce the protease, as the diameter of the hemolysis area on the test medium reached about 22.16 mm, and the lipase, as the hemolysis area reached 15.18 mm, and the production of the urease was 100%. Regarding the drug sensitivity of A. chevalieri to antifungals utilizing the disk diffusion technique, Nystatin proved to be the most effective in inhibiting fungal growth with an average diameter of 25.3 mm, while Amphotercin B and Itraconazole gave a positive result but less inhibition than Nystatin, and their inhibition percentage was 10.5 mm and 18.1 mm, while the result of the test for Fluconazole was negative and did not provide effectiveness in inhibiting fungal growth.

Design for The Purpose of Protecting Encryption Systems and Cloud Database Files

2026-03-01T00:20:19+00:00

Cloud database security might be an essential question for cloud databases since enterprises store and process huge volumes of sensitive data in the cloud. In several situations, cryptosystems were essential to maintaining the data’s integrity and secrecy. This study has comprehensively investigated the application of cryptosystems for the cloud database security. The paper first discusses the premise and importance of cloud database security, with a focus on the increasing demand for advanced information security schemes. The literature is critically examined in terms of major achievements and challenges in this field. Following this, we present a study on the application of cryptosystems to cloud databases, including their trade-offs and execution. It also discusses possible trends and developments on these lines for improving cloud database security, such block chain, holomorphic encryption, multi-factor authentication. Furthermore, the paper identifies several questions with regard to possibilities and future directions in the area encompassing novel encryption schemes, execution optimization, key management, and usability enhancements, regulatory issues, as well as integration with emerging technologies. Also, we have proposed a encryption based file security software using. The proposed solution can be adapted to protect files in cloud database security. The study ends with the implication that the significance of cryptosystems for cloud database security, such as in the case of this research, should be a priority in future research and development efforts to overcome constraints and stimulate growth in this area.

A Comprehensive Review of Privacy-Preserving Techniques in Artificial Intelligence and Machine Learning: Challenges, Solutions, and Future Directions

2026-03-01T00:48:49+00:00

Fast progress of new technologies like Artificial Intelligence (AI) and Machine Learning (ML) is bringing forth important issues when it comes to data privacy in different fields. AI and ML solutions depend on great volumes of data that frequently include personal or organizational data, some of which is sensitive. Preserving privacy with no degradation of model performance presents major technical, ethical and legal issues. This survey presents an extensive overview of the most widely adopted privacy-preserving methodologies in AI and ML, including encryption based techniques, federated learning, differential privacy, data anonymization, blockchain-inspired approaches as well as private AI APIs and synthetic data. 875 Each approach is evaluated by its respective advantage, disadvantage and applicable conditions. Moreover, a comparison shows trade-offs between security level, performance and scalability and hybrid solutions are found most promising for practical use cases. Finally, the paper concludes by providing a few research directions, in particular highlighting the mandatory support for adaptive privacy preservation measures, scalable solutions and consistent global legislations. This is a thorough reference text for those who wish to design secure, performant and privacy-focused AI and ML systems.

Optimized Deep Neural Network to Optimally Classify Attacks as an Intrusion Detection System

2026-03-02T09:56:36+00:00

The expansion of internet usage has made networks increasingly vulnerable to continuous and escalating breaches. This is due to the critical nature of the information exchanged across these networks. Given the importance of this information, a method for protecting it from breaches is essential. Several such methods exist, including those within machine learning (ML) and deep learning (DL). In this research, we will utilize the Genetic Algorithm (GA) to determine the optimal hyperparameter values for the deep learning model, which will then be used to classify attacks. Results show that the accuracy was comparable to other works, 98.54%, with a precision of 98.55%. The proposed deep neural network is based on convolutional neural network (CNN). On the other hand, the suggested model will need, of course, to a training/testing dataset, thus, the NSL-Kdd dataset was affordable, where it was cleaned and prepared for the training and testing purposes. Last but not least, it is recommended to make use of the suggested approach to be generalized against attacks in real-world systems.

Review on Triple Integral Transforms and Their Inverses

2026-03-02T14:10:39+00:00

Transformations involving triple-integrals are the basis for the solutions of a number of three-dimensional partial differential equations that arise in physics, engineering, and applied sciences. Although many triple transforms are available (e.g., Laplace, Fourier, Aboodh, Shehu, Ezaki, and various hybrid triple transforms), to the best of our knowledge, a detailed systematic comparative review is not yet available that compares different triple transforms with respect to their definitions and kernels, inverse formulations, and applications. This paper fills this gap by giving an overview of most important, widespread triple integral transforms, whereby introducing their formal definitions, inverse transforms as well as structural properties. Analytical behaviors of benign configurations are presented in solving different problems such as heat, wave, diffusion, and boundary value problems, which differ due to differences in kernel structures, convergence conditions, and their parameter configuration. We show that all the triple transforms impose preservation of linearity and algebraic simplification of differential operators, but each possess respective benefits specific to final transformed domain characteristics and boundary conditions. The results highlight the significance of careful selection of an adequate transform based on the structural characteristic of the problem under investigation and that future research should include generalized and hybrid triple transforms coupled with numerical methods to provide a more time-efficient analytical–numerical model

Intelligent Crop Prediction and Smart Irrigation System Using IoT, K-Means Clustering, and Random Forest Algorithms

2026-03-03T10:41:27+00:00

Agriculture covers more than 60.43% of the land in India. But we can't meet the country's growing needs if we keep using old farming methods. We need to sustainably and efficiently harvest most of the available farmland. The goal of this project is to set up a cost-effective system that uses high-tech sensors and the Internet of Things to make crop harvesting easier. The most important parts of the system are that it is a module that can be controlled by a mobile app and uses a set of sensors (electromagnetic, NPK, optical, and electrochemical) to measure the soil's texture and nutritional content. Advanced K-Means Clustering, Random Forest, and Decision Tree algorithms are used to analyse it and guess which crop can be grown. It uses infrared and laser sensors to figure out the best way to plant seeds so that they grow the most. It also keeps an eye on how wet the soil is and sets up irrigation systems like drip irrigation. A built-in weather forecasting system will predict rain and change the irrigation cycle to lower the risk of over- or under-irrigation. The module alerts the farmer when the crop needs something during the growth phase. It also has a flexible infrared motion sensor to detect movement and an alarm system to keep predators away, which is another way to keep it safe. Mobile apps control all the functions, while Machine Learning algorithms manage and oversee them. High-tech sensors and actuators, along with microcontrollers and Raspberry Pi, carry out the tasks.

Newton–Kantorovich Based Numerical Methods for Nonlinear Integral Equations

2026-03-01T07:32:10+00:00

This research addresses the development of an effective numerical method for solving nonlinear Volterra integral equations of the second kind based on the Newton–Cantorovich method for bypassing nonlinearity and converting the problem into a series of successive linear equations. This technique is combined with an adaptive trapezoidal rule to improve the accuracy of numerical integration by adjusting the step size according to the behavior of the function. This combination contributes to reducing cumulative error and enhancing convergence stability during iteration. The method was implemented using MATLAB and tested on several standard examples. The results showed high agreement with exact solutions and significant improvement compared to traditional methods such as Simpson's rule, confirming the efficiency and computational accuracy of the proposed method.

Enhancing Security in Software-Defined Networking: A Framework for Encryption

2026-03-05T11:08:30+00:00

Software-Defined Networking (SDN) will provide a higher network controllability and flexibility through decoupling of the control and data planes. Nonetheless, its centralized design presents security vulnerabilities and thus it is susceptible to threats like unauthorized access, data breaches and controller attacks. A secure SDN environment is achievable by having strong encryption, authentication and access control. Some of the protective measures that have been applied in this work are logging and security audit services, integration of the SSL/TLS and enforcing authentication of the graphical user interface (GUI). Also, encryptions are used based on cryptographic cipher like DES, AES and role-based authorization is done with the help of FortNOX to improve access controls. The suggested security model enhances the SDN controller by reducing possible threats and ensuring the entire system resiliency. Encryption methods coupled with access control measures provide a better level of data confidentiality and integrity lowering threats in SDN communication. These steps will help make SDN an environment that is more reliable.

Computational Simulation of the Lotka–Volterra Predator–Prey Model

2026-03-08T04:31:28+00:00

Predator-prey interactions constitute a fundamental pillar of ecological theory,
characterizing the periodic fluctuations between consumer populations and their resources. This
research utilizes the Lotka–Volterra framework to analyze system stability and oscillatory dynamics
across a spectrum of environmental variables. By employing advanced numerical methods to
resolve a system of nonlinear differential equations, the study facilitates a precise visualization of
population trajectories. Temporal analysis of the simulation data reveals a consistent rhythmic
displacement, where surges in prey abundance act as a leading indicator for predator growth—a
classic phase-lag relationship. These findings highlight the indispensable role of computational
modeling in decoding the regulatory mechanisms of population cycles. Beyond theoretical ecology,
the versatility of this framework offers a robust simulation environment for diverse applications,
ranging from agricultural pest management and wildlife conservation to the modeling of
epidemiological and therapeutic interventions in medical science.

Blockchain-Based Federated Learning for Privacy-Preserving AI in Smart City

2026-03-11T07:57:30+00:00

Artificial Intelligence (AI) becomes more and more common in infrastructures of Smart Cities with the use of AI to optimize traffic, manage energy consumption, and monitor the environment. There are however major concerns that come with centralized AI training such as privacy risk, governance of data, and single points of failure. The paper introduces a blockchain-enabled federated learning (hereinafter, BFL) framework to organize the training of AI models to ensure privacy and decentralization of training processes on heterogeneous internet of things (IoT) devices located in the urban setting. The framework uses a lightweight Proof-of-Authority (PoA) blockchain to securely, tamper-proof aggregate and log transparent participation and actively uses adaptive compression techniques in model communication and storage costs. An incentive mechanism can be set up with the help of a smart contract to welcome a wider range of stakeholders in the city. Experimental analysis conducted on the METR-LA traffic dataset shows that BFL can attain similar performance evaluation to the centralized methods whilst having lower privacy leakage risks, possessing low blockchain latency (<200 ms), and saving 45 percent in communication expense. The suggested framework gives a scalable and trustworthy AI training scheme to intelligent cities.