This paper introduces an enhanced YOLOv11 model for accurate and efficient weed detection in precision agriculture. Our improvements to YOLOv11n include integrating SPD-Conv for faster speed and better multi-scale fusion, embedding the SCSA attention mechanism in Bottlenecks to mitigate occlusion-related misses, and utilizing the CIoU loss for an improved precision-recall balance. Leveraging the CottonWeedDet12 dataset augmented with random transformations, our model achieves a 1% higher mAP than YOLOv11n and outperforms counterparts like YOLOv5/7/8. It demonstrates robust performance across varying lighting, occlusion, and complex soil conditions, offering a viable solution for smart farming applications.

The paper provides an overview of the anomaly detection techniques to address the ever-evolving landscape of code vulnerabilities. Main objective is to provide a comprehensive understanding of the various methodologies, algorithms, and frameworks that have been developed to enhance the security and reliability of software systems. In paper determines early research in anomaly detection techniques from 2019 and followed methods, approaches and frameworks until recent times. According to research are defined challenges such as high computational costs, false positives, and the need for large, labeled datasets persist. The work specifies ML methods and calculates classic classification metrics such as accuracy, F1 score, precision, recall, geometric mean, etc. This review underscores the importance of continuous advancements in anomaly detection techniques to address the ever-evolving landscape of code vulnerabilities.

This paper examines the combination of traditional banking credit assessment techniques with contemporary internal sales accounting systems in Kazakhstan, aiming to augment the precision and resilience of financial assessments pertaining to SMEs. The proposed model consists of two discrete components: a traditional credit scoring module that employs logistic regression and a supplementary sales analytics module that leverages ensemble machine learning methodologies — random forests and gradient boosting algorithms. The outputs generated by these components are amalgamated through an ensemble strategy, where optimal weighting coefficients are ascertained via cross-validation. An empirical analysis was conducted on a dataset encompassing 41,000 SME records from a prominent Kazakhstan bank alongside daily transactional sales data from 150 SMEs gathered between the years 2021 and 2024. The integrated hybrid model demonstrated a statistically meaningful enhancement in predictive efficacy, as evidenced by an increase in the area under the ROC curve from 0.76 to 0.87 and a decrease in mean squared error from 0.12 to 0.08 relative to the traditional methodology. The investigation delves into the transformative influence of digitalization on innovation within SMEs, elucidating that improved real-time data integration not only sharpens risk assessment processes but also promotes adaptive lending strategies and operational efficiencies.

The development of credit scoring is one of the key topics of attention in credit risk management in financial companies. However, a single approach to produce rating cards is frequently worthless since loan products differ in risk and financing time and often there is insufficient information on borrowers. The paper addresses the features of creating score cards for consumer credit, refinancing, small and medium businesses, auto loans, mortgage loans, fintech and P2P lending. Thus, the present work can be considered as the above comparative analysis of the most important elements influencing the probability of default of the borrower in the settlement by segments, together with the consideration of machine learning techniques and the use of alternative data sources that can improve the accuracy of the forecast. Depending on the usual credit product, the analysis lets one create recommendations for choosing the optimal approach of creating scoring cards, so enhancing the accuracy of the borrower’s creditworthiness projection and reducing the degree of default risk.

Spatiotemporal analytics of population movement and density data plays a crucial role in building a «smart city», providing a basis for optimizing urban planning, improving transport systems, increasing public safety, environmental monitoring, developing digital services and urban analytics. This article presents the results of a study on spatiotemporal patterns of distribution and concentration of the population of Almaty using the method of dynamic heat maps. To build a complete picture of the movement, density and activity of the population, open geographic data from OpenStreetMap (OSM) and aggregated data from a mobile operator were used. Analysis of the load on urban quadrants of 500×500 meters based on OSM made it possible to assess the key patterns of change in population density depending on the time of day. Visualization of spatiotemporal data is implemented using the Python Folium library, which ensured the creation of clear interactive maps. The scientific novelty of the study lies in the study of urban processes in Almaty based on integrated data from different sources reflecting the spatiotemporal features of the dynamics of the urban population. The results obtained demonstrate clear patterns of population concentration that can be used to more accurately forecast and plan the allocation of resources and urban infrastructure.

This paper introduces a hybrid credit assessment methodology tailored specifically to the unique financial, informational, and operational challenges faced by importers. Importers operate at the intersection of cross-border financial risk, fluctuating supply chains, and opaque transactional histories - factors that challenge conventional credit models. While banks rely on structured financial records and established credit histories, P2P platforms harness alternative data and decentralized trust mechanisms such a peer feedback and real-time behavioral signals. Rather than merely merging technologies, the proposed approach integrates differing capabilities: institutional data interpretation dynamical risk modeling, and platform-based investor participation. This synthesis enables a more adaptive, inclusive, and context-aware assessment process. Core components include continuous data validation, alternative trade-data integration, machine learning-driven decision support, and modular feedback loops from lender communities. The methodology is evaluated through a structured literature review and illustrative case applications. By focusing on the informational volatility and operational complexity unique to import-driven businesses, this research contributes a targeted solution to a global financing gap - advancing both credit innovation and trade accessibility.

This article presents a comparative analysis of international benchmarking systems and their application to assessing urban livability and citizen engagement. The research examines key global indices – Economist Intelligence Unit (EIU), Mercer, PwC, TUWIEN Smart City Model, MIT Treepedia, and the National League of Cities (NLC) – to identify dominant trends and classification models of city benchmarking. Four major types of benchmarking practices were defined: multi-factor indices, single-indicator rankings, thematic analytical reviews, and diagnostic metrics. An empirical case study of Almaty, Kazakhstan, demonstrates the adaptation of international practices to the local context. Between 2020 and 2023, over 1,500 participatory projects were implemented under the Participatory Budget program, primarily in urban greening, infrastructure, and public safety. The findings show that digital governance platforms (Open Almaty, iKomek, Almaty Urban Center) enhance civic participation but remain limited by unequal digital access. The study concludes that benchmarking serves as an effective governance and evaluation tool for improving urban livability and inclusiveness. The Almaty case illustrates the potential of combining global best practices, data-driven governance, and participatory approaches to foster sustainable urban transformation. © 2025, Kazakh-British Technical University. All rights reserved.

Gene expression analysis has become a key component in understanding cellular behavior, disease mechanisms, and drug response. The advent of high-throughput sequencing, particularly single-cell RNA sequencing (scRNA- seq), has expanded our ability to study cellular heterogeneity to an unprecedented level. Clustering algorithms needed to group genes or cells with similar expression profiles have become invaluable for analyzing the massive data sets generated by these technologies. This article reviews various clustering methods applied to gene expression data, particularly single-cell RNA sequencing. The analysis covers traditional methods such as hierarchical clustering and k-means, as well as more advanced approaches such as model-based clustering, machine learning-based methods, and deep learning methods. The primary challenges encompass handling high-dimensional data, mitigating noise, and achieving scalability for large datasets. Moreover, new advancements such as multi-omics data integration, deep learning-based clustering, and federated learning offer potential enhancements in accuracy and biological relevance for clustering applications in gene expression research. The review concludes with a discussion of clustering algorithms in handling increasingly complex gene expression data for more accurate biological insights.

With the rapid development of big data and internet technology, big data financial platform companies collect and organize massive data through their own platforms, improve credit scoring parameters, and use machine learning methods to conduct comprehensive and scientific credit scoring assessments. Thus, banks in the construction of credit scoring face big problems. Based on the limitations of the existing system and methods of personal credit score, it is necessary to study personal credit score based on machine learning methods, improve the parameters and scoring system of personal credit score, clarify data collection channels, and use dynamic desensitization technology to desensitize data, LOF test method to test outliers' data and a random forest method to complete missing data values. You then use the gradient boosting decision tree method to view the important indicators, process the tested indicators with a scorecard model based on logistic regression, and derive a personal credit score. Finally, the model is tested with a BP neural network and the model is used to predict the level of personal credit. The study shows that machine learning can further improve the accuracy of individuals' credit ratings and provide the scientific basis and reference for credit ratings of commercial banks.

The object of the study is the text classification and semantic search tailored to the unique linguistic features of the Kazakh language. The research addresses the challenge of improving the accuracy, relevance, and efficiency of semantic search. This study focuses on improving semantic search for the Kazakh language by analyzing computational models tailored to its unique linguistic features, such as agglutinative morphology and rich inflectional systems. The research compares traditional rule-based approaches and advanced transformer architectures, including fine-tuned models like RoBERTa, for their ability to handle semantic nuances, contextual relationships, and user intent. The results reveal that fine-tuned transformer models achieved significant advancements, with the RoBERTa model attaining a Precision@10 of 89.4 %, a Mean Reciprocal Rank (MRR) of 85.6 %, and an F1-Score of 88.0 %. Additionally, the semantic search system developed in this study demonstrated a precision of 88.4 %, recall of 87.6 %, and an F1-score of 88.0 % on a domain-specific Kazakh dataset. Key to these improvements were innovations in preprocessing pipelines, including custom tokenization and lemmatization tailored to Kazakh's agglutinative morphology, and the integration of contextual embeddings to resolve issues such as synonymy and homonymy. Computational efficiency was enhanced through resource optimization techniques, enabling the deployment of these advanced models in constrained environments. These findings underscore the potential of tailored transformer models to bridge the gap in semantic search capabilities for underrepresented languages like Kazakh, advancing the inclusivity of natural language processing technologies