AUTOMATIC CONTROL AND COMPUTER SCIENCES最新文献

{"title":"Optimizing Credit Card Fraud Detection Using Ensemble Methods and Advanced Oversampling Techniques","authors":"Jasvant Mandloi,&nbsp;Harshita Mandloi,&nbsp;Rakesh Kumar Bhujade","doi":"10.3103/S0146411625602990","DOIUrl":"10.3103/S0146411625602990","url":null,"abstract":"<p>Fraud incidents involving credit cards pose a persistent and escalating threat to financial institutions and consumers globally, necessitating the development of sophisticated and efficient detection mechanisms. This study examines the effectiveness of contemporary machine learning methods in identifying fraudulent credit card transactions, addressing the prevalent issue of class imbalance in fraud datasets (284,807 transactions; only 0.17% fraudulent). Utilizing the publicly accessible Kaggle Credit Card Fraud dataset, we employed and evaluated eight supervised learning models—logistic regression, support vector machine, random forest, XGBoost, gradient boosting machine (GBM), voting classifier, artificial neural network (ANN), and convolutional neural network (CNN). SMOTE (synthetic minority over-sampling technique) was applied to address class imbalance. Each model was evaluated on Accuracy, Precision, Recall, F1-score, and ROC-AUC. The voting classifier (XGBoost + GBM + random forest) achieved the highest overall performance with Accuracy = 99.12%, Precision = 97.10%, Recall = 89.89%, F1-score = 93.37%, and ROC-AUC = 0.98. XGBoost independently delivered competitive results with ROC-AUC = 0.97 and Recall = 87.65%, demonstrating its strength in handling imbalanced data. SMOTE markedly enhanced detection rates by equilibrating the dataset and facilitating more representative learning. These results confirm that ensemble learning integrated with advanced resampling techniques offers an accurate, scalable, and affordable architecture for real-world fraud prevention systems, reinforcing the burgeoning field of financial cybersecurity.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 3","pages":"363 - 376"},"PeriodicalIF":0.5,"publicationDate":"2026-10-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149219417","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SHA-ELM-IPS: A Fast and Robust Intrusion Prevention Framework with Hash-Based Blocking for Modern Networks","authors":"Sarika,&nbsp; Rajeshwar Dass","doi":"10.3103/S0146411626700501","DOIUrl":"10.3103/S0146411626700501","url":null,"abstract":"<p>This research article presents a novel intrusion prevention system (IPS). It is a completely lightweight and high-performance along with a hashing based on SHA-256 method and a classifier based on extreme learning machine (ELM) for real-time cyber defense. Besides, proposed framework combines signature-based detection by using hash matching and fast learning abilities of ELM for efficient detection and second by second blocking of malicious traffic without adding any computational overhead. The system was tested on two very popular benchmark datasets, NSL-KDD and CIC-IDS-2017, to check its robustness and scalability. The experiments show that the new model is 96.73 and 99.28% accurate on the NSL-KDD and CIC-IDS-2017 datasets respectively, precision, recall and F1-scores are also high exceeding 96 and 99%. Furthermore, the model produces very good ROC-AUC scores of 0.9665 and 0.9923, which are indicators of great classification power. In fact, the system is not only designed to minimize the number of false positive cases but also to keep very high values of the Matthews correlation coefficient (MCC), so the detected cases are trustworthy. Also, the IP blocking at the network layer is an excellent preventive measure against hostile entries as it records blocking percentages of 45.91 and 57.27%. Furthermore, the system has short training and testing duration that make it an especially useful tool for being actively present in the field. Basically, the SHA-ELM-IPS growth can be a scalable, productive, and most of all, a very realistic solution for stopping intrusions in the continuously changing network settings.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 3","pages":"317 - 324"},"PeriodicalIF":0.5,"publicationDate":"2026-10-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149219485","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Warning Model of Credit Default Risk Based on Hierarchical Clustering-Neural Network","authors":"Zhanjiang Li,&nbsp;Tongtong Wang,&nbsp;Qinjin Zhang","doi":"10.3103/S0146411626700392","DOIUrl":"10.3103/S0146411626700392","url":null,"abstract":"<p>The early warning of default risk is an important issue closely concerned by all stakeholders of family farms and ranches (FF&amp;R). However, due to the lack of current research on credit risk evaluation of FF&amp;R, the existing credit default risk early warning models of FF&amp;Rs mostly follow the credit default risk early warning model of farmers and small and micro enterprises. Therefore, the construction of credit default risk early warning model of FF&amp;R is both a theoretical problem to be solved and of great practical significance. This paper classifies FF&amp;Rs into three types of nondefault, low default and high default by using hierarchical clustering, so as to obtain a more realistic description of default conditions of FF&amp;Rs. At the same time, the classification of three default states is taken as the input layer of the neural network, and the number of selected indicators of FF&amp;Rs is taken as the output layer to construct the default risk warning model of FF&amp;R neural network, thus improving the fitting efficiency of the neural network and the accuracy of default prediction state, timely warning and preventing the further deterioration of the default situation of FF&amp;R. Empirical evidence shows that the hierarchical clustering-neural network model constructed in this paper achieves a default prediction accuracy of 97.9% on the FF&amp;Rs dataset. Even after K-fold cross-validation, the model’s accuracy remains at 90.66%. Furthermore, compared with five traditional evaluation models, the model in this paper still maintains a high level of accuracy.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 2","pages":"133 - 142"},"PeriodicalIF":0.5,"publicationDate":"2026-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149023035","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Interactive Smooth Variable Structure Filter and Its Application in Integrated Navigation","authors":"Lei Wang,&nbsp;Lin Fang,&nbsp;Mingyu Zhang","doi":"10.3103/S0146411626700471","DOIUrl":"10.3103/S0146411626700471","url":null,"abstract":"<p>Addressing the challenges of time-varying and non-Gaussian noise in inertial navigation system (INS) and Global Positioning System (GPS) integrated navigation, traditional methods often suffer from limited robustness and accuracy due to their dependence on precise noise statistics and system models. To overcome these limitations, an improved interacting multiple model smooth variable structure filter algorithm is proposed in this paper. The novelty of our work lies in the deep integration of an improved smooth variable structure filter (ISVSF) with the interacting multiple model (IMM) framework. Based on sliding mode control and variable structure theory, the smooth variable structure filter (SVSF) confines state estimates within a bounded region around the true state trajectory, offering enhanced robustness to model uncertainties compared to the Kalman filter. Building on SVSF, the proposed ISVSF incorporates a time-varying smooth subspace with a smooth saturation function for gain calculation and uses Bayesian theory to derive the covariance matrix, thereby improving filtering stability and convergence. The IMM algorithm, recognized as an effective approach for state estimation in systems with variable structures and parameters, helps mitigate navigation parameter errors induced by time-varying noise. The actual field experiments show that the proposed IMM-ISVSF algorithm can improve the estimation accuracy of attitude, velocity and position of the integrated navigation system compared with the smooth variable structure filter and the interactive multimodel extended Kalman filter.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 2","pages":"246 - 259"},"PeriodicalIF":0.5,"publicationDate":"2026-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149023261","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"APBROP: Adaptive Priority-Based Routing Protocol for Enhanced Performance in Manets","authors":"Sonia Singhal,&nbsp; Ashwani Kush","doi":"10.3103/S0146411626700446","DOIUrl":"10.3103/S0146411626700446","url":null,"abstract":"<p>Mobile ad hoc networks (MANETs) play a critical role in facilitating communication in dynamic, infrastructure-less environments such as disaster management scenarios, military operations, and remote area networking. However, traditional routing protocols like AODV and OLSR face significant challenges, including increased latency, inefficient route discovery, and poor adaptability to frequent topology changes. This paper introduces the Priority-Based Routing Optimization Protocol (APBROP), a novel hybrid routing framework that addresses these limitations through a combination of innovative techniques. APBROP employs a dynamic multimetric weighting mechanism that integrates; hop count, bandwidth, and link stability to ensure optimal route selection. It incorporates a priority-based route discovery system to prioritize critical data traffic to reduce network congestion and extend node lifetimes. Additionally, APBROP leverages a cluster-assisted reactive-proactive mechanism, which combines intracluster proactive routing with intercluster reactive routing, to enhance scalability and reduce control overhead. To improve resilience, the protocol includes a dynamic failure recovery system that preemptively detects and mitigates route failures using multipath redundancy. Simulation results demonstrate that APBROP significantly outperforms existing protocols in terms of packet delivery ratio, end-to-end delay, throughput and scalability, particularly in disaster-prone and resource-constrained environments. These results establish APBROP as a robust, adaptable, and setting a new benchmark for MANET routing protocols.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 2","pages":"208 - 219"},"PeriodicalIF":0.5,"publicationDate":"2026-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149023039","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Analysis of the Application Effect of Machine Vision Equipment in the Analysis of the Moment of Inertia Measuring System","authors":"Su Yang,&nbsp;Xiu Hang,&nbsp;Wang Li Fang,&nbsp;Cheng Lei,&nbsp;Qin Qi","doi":"10.3103/S0146411626700483","DOIUrl":"10.3103/S0146411626700483","url":null,"abstract":"<p>This paper proposes a machine-vision-based moment-of-inertia (MOI) measurement system that balances cost and accuracy. It features an annular diagonal coding landmark and a recognition algorithm for precise identification, real-time tracking, and multipoint localization. By integrating vision algorithms with the torsion pendulum method, the system achieves high accuracy with minimal error. Experimental results show a standard deviation of 1.79 × 10^–5 kg m², confirming repeatability. Additionally, damping parameters and torsional vibration period help mitigate damping effects, enhancing measurement reliability.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 2","pages":"272 - 285"},"PeriodicalIF":0.5,"publicationDate":"2026-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149023262","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Optimal Control of Completely Unknown Nonlinear Systems Based on SDRE Method","authors":"Xueping Xiong,&nbsp; Xiaohua Wang","doi":"10.3103/S0146411626700410","DOIUrl":"10.3103/S0146411626700410","url":null,"abstract":"<p>This paper proposes a novel online iterative algorithm for optimizing the control of nonlinear systems, enabling optimization of continuous nonlinear systems with unknown dynamics. The method builds upon traditional state-dependent Riccati equation (SDRE) concepts, segmentally approximating and linearizing the nonlinear system, transforming the Hamilton–Jacobi–Bellman (HJB) equation into a solvable SDRE problem. By applying an arbitrary stable control and utilizing real-time data, this method eliminates the need for system matrices A and B, directly solving the optimal control problem through a data-driven approach. Strategy iteration is employed to approximate the solution of the Riccati equation for each linearized system segment, which is then applied in real-time without requiring prior knowledge of the system’s dynamics. Unlike conventional SDRE-based algorithms that rely on accurate system modeling, the proposed method remains effective even when the system model changes, addressing a key limitation of traditional approaches. The paper also presents a convergence proof for the recursive algorithm and validates its effectiveness through simulations on second-order affine nonlinear continuous systems. Compared to traditional SDRE control methods, the proposed approach offers substantial advantages in adaptability and performance.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 2","pages":"164 - 175"},"PeriodicalIF":0.5,"publicationDate":"2026-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149023038","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Dynamic BI-Handed Sign Recognition Using 3D-CNN and Gated Recurrent Unit for Indian Sign Language","authors":"Syed Farhan Naqvi,&nbsp;Smeet H. Ramteke,&nbsp;R. Sreemathy,&nbsp;Jayashree Jagdale,&nbsp;Aagaaz Ali Sayed","doi":"10.3103/S014641162670046X","DOIUrl":"10.3103/S014641162670046X","url":null,"abstract":"<p>Sign Language recognition systems play a pivotal role in fostering inclusivity and accessibility for the hearing and speech impaired. In this paper, we present a novel approach to Indian Sign Language recognition (ISL) recognition using deep learning techniques. Leveraging a self-created comprehensive dataset of dynamic ISL gestures, each represented as videos, created by volunteers under the supervision of experts, we propose a robust model architecture tailored to the unique characteristics of ISL. The emergence of convolutional neural networks (CNNs) and sequence-based models over the past decade has propelled the advancement of Sign Language Recognition (SLR) through artificial intelligence (AI). However, the domain of Indian Sign Language (ISL) has remained relatively under explored, primarily due to the absence of a comprehensive open-source dataset. Addressing this gap, our research endeavours to fill this void by conducting a thorough investigation and implementation of spatial and temporal feature-based methodologies on a newly created dataset. We utilized Mediapipe to extract spatial features. These features were then seamlessly integrated into a sequence-based gated recurrent unit (GRU) model, ensuring a robust framework for analysis. For the temporal dimension, we adopted a sophisticated sliding window approach, facilitating the training of three distinct convolutional neural network (CNN) models on a subset encompassing 21 dynamically bi-handed signed words. This work delineates the architectural intricacies underpinning each methodological approach, complemented by a comprehensive exposition of our newly recorded dataset. On the stated dataset and GRU technique using key-point method, we achieved the accuracy of 99.71%. Furthermore, we present an in-depth comparison of the accuracy attained by each technique employed, contributing to a nuanced understanding of their respective capabilities within the domain of Indian Sign Language recognition.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 2","pages":"231 - 245"},"PeriodicalIF":0.5,"publicationDate":"2026-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149023034","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Optimization of Lidar SLAM Positioning Using Error Iterative Extended Kalman Filtering with Multisensor Fusion","authors":"Bo Chen,&nbsp;Jianjun Wang,&nbsp;Weirun Song,&nbsp;Xiaoxiao Cheng,&nbsp;Kun Wang,&nbsp;Xudong Li","doi":"10.3103/S0146411626700434","DOIUrl":"10.3103/S0146411626700434","url":null,"abstract":"<p>The traditional LiDAR SLAM map construction and real-time positioning methods usually use the extended Kalman filter method. However, in outdoor scenes, on the one hand, due to the unclear environmental characteristics, the laser point cloud pairing error is large. On the other hand, the nonlinear calculation of the extended Kalman filter has accumulated errors, resulting in large positioning errors for robots and low map construction accuracy. Two measures have been taken to improve the reliability of robot spatial positioning. Firstly, based on the laser odometer in LiDAR SLAM, three sensors, namely photoelectric encoded odometry, GPS (Global Positioning System), and IMU (inertial measurement unit), are integrated to improve the reliability of robot spatial positioning. Secondly, a multisensor fusion extended Kalman filtering algorithm based on error state iteration calculation is proposed to further compress the error space. Firstly, a kinematic model of the wheeled mobile robot was constructed, along with spatial measurement and positioning models for four sensors: photoelectric odometry, GPS (Global Positioning System), IMU (inertial measurement unit), and IMU. Secondly, based on the IMU spatial positioning error state iterative calculation model, the extended Kalman filter fusion algorithm is adopted to predict and iteratively update the spatial pose prediction data of the four sensors, effectively improving the spatial positioning accuracy of the robot. Finally, a physical experimental platform for unmanned wheelchair LiDAR SLAM was established, and experimental verification was carried out on the campus environment. The results showed that the robot trajectory positioning error of the LiDAR SLAM optimized based on the proposed error iterative multisensor fusion extended Kalman filter algorithm was only 0.040 m, which reduced the trajectory error by 0.017 m compared to other fusion methods and improved the accuracy by 29.82%. At the same time, the accuracy of the constructed 3D point cloud map was significantly improved.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 2","pages":"191 - 207"},"PeriodicalIF":0.5,"publicationDate":"2026-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149023037","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Complex Data Linkages in Buildings Can Be Modeled and Visualized Using Bayesian Networks","authors":"Pengcheng Wang,&nbsp;Yuqi Bing,&nbsp;Huan Zhou,&nbsp;Hai Liu,&nbsp;Meilingzhi Chen","doi":"10.3103/S0146411626700458","DOIUrl":"10.3103/S0146411626700458","url":null,"abstract":"<p>As smart cities continue to grow quickly, smart buildings, one of their main constituents, are crucial to raising both the general competitiveness of cities and the standard of living for their citizens. By dynamically mapping and simulating building architecture, digital twin-based smart building modeling offers data support throughout the entire life cycle, from design to operation. When it comes to modeling and displaying complex data interactions in buildings, the existing technology still faces obstacles, including difficult model design, low data processing efficiency, and inadequate real-time performance. Because of this, this research suggests a Bayesian network-based approach for modeling and displaying intricate data interactions in structures. In addition to expressing the conditional independence between building data, the Bayesian network, a probabilistic graphical model, facilitates two-way reasoning from cause to effect and from effect to cause. The Bayesian network facilitates the intuitive modeling and reliability assessment of complex building systems by mapping different kinds of building state variables and their logical relationships into nodes and conditional probability distributions when combined with the fault tree analysis method. Additionally, a digital twin building visualization platform that facilitates real-time interaction is built by incorporating Unreal Engine’s high-performance rendering technology, which significantly enhances the user experience and expressiveness of data modeling. In terms of data processing efficiency, model accuracy, and real-time performance, the experimental results demonstrate that the Bayesian network-based modeling approach outperforms the current technology. This offers theoretical backing and a technical guarantee for the intelligent management of smart buildings.</p>","PeriodicalId":46238,"journal":{"name":"AUTOMATIC CONTROL AND COMPUTER SCIENCES","volume":"60 2","pages":"220 - 230"},"PeriodicalIF":0.5,"publicationDate":"2026-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149023260","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}