Procedia Computer Science最新文献

{"title":"The Construction of Generative Artificial Intelligence Algorithms and Neural Network Models for Personalized Learning","authors":"Sai Zhang,&nbsp;Qing Han,&nbsp;Lingzhi Song","doi":"10.1016/j.procs.2026.03.294","DOIUrl":"10.1016/j.procs.2026.03.294","url":null,"abstract":"<div><div>Traditional educational models suffer from low knowledge coverage and a lack of quantifiable evaluation when generating personalized learning paths. This paper employs an improved Transformer-XL architecture as the core generative model, incorporating a knowledge graph triple embedding module to achieve semantic enhancement of subject knowledge nodes. Through a gating mechanism, knowledge graph features are dynamically integrated with sequence generation features to optimize the model’s ability to capture logical relationships in educational contexts. Experiments show that the knowledge coverage of the generated paths averages 92.4%, the path coherence dimension ranges from 0.88 to 0.95, and the completion error is reduced, demonstrating that the model has significant advantages in ensuring knowledge integrity and behavioral feasibility.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"279 ","pages":"Pages 801-808"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148182873","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":"Patient Risk Assessment in Emergency Departments using Ensemble Learning","authors":"Mehanas Shahul,&nbsp;Pushpalatha K P","doi":"10.1016/j.procs.2026.05.130","DOIUrl":"10.1016/j.procs.2026.05.130","url":null,"abstract":"<div><div>Efficient patient allocation within a hospital’s Emergency Department (ED) is crucial in ensuring prompt and appropriate medical intervention, especially under high patient load conditions. In this study, we developed an ensemble-based machine learning framework to predict patient risk status during triage, using a newly curated medical dataset consisting of fundamental vital sign features such as heart rate, blood pressure, temperature, respiratory rate, and oxygen saturation (SPO2). To overcome the inherent class imbalance problem commonly observed in clinical data, the Synthetic Minority Oversampling Technique (SMOTE) was employed, which significantly improved model stability and predictive accuracy. Various ensemble algorithms were implemented and compared, including Gradient Boosting Classifier (GBC), Light Gradient Boosting Machine (LGBM), XGBoost (XGB), and Extra Trees Classifier (ETC). Among these, the Gradient Boosting Classifier exhibited superior overall performance in accuracy (87.21%), recall (60.12%), F1 score (67.11%), and AUC (77.42%), demonstrating its robustness and reliability for medical risk prediction. The GBC Classifier further achieved the highest sensitivity (60.12%), indicating its strong ability to correctly identify high-risk patients while maintaining excellent specificity (94.72%) for accurately classifying low-risk cases. This approach represents a promising advancement toward intelligent, automated risk-level assessment systems that can improve decision-making and enhance the judicious allotment of resources in emergency healthcare settings.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"282 ","pages":"Pages 862-871"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148574608","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Robust Lightweight Authentication and Key Establishment Protocol for Internet of Drone Networks","authors":"Ram Prakash,&nbsp;Vinod Kumar,&nbsp;Khushboo Gupta","doi":"10.1016/j.procs.2026.05.129","DOIUrl":"10.1016/j.procs.2026.05.129","url":null,"abstract":"<div><div>The internet of drones (<em>IoD</em>) is a growing technology utilized in a diverse range of applications, including surveillance data and images sharing, smart cities, industrial sectors, and transportation systems. In order to guarantee secure access to real-time data from drones, numerous researchers have developed various protocols for authentication and key agreement. However, there are a number of security and performance issues with these protocols. In this paper, we introduce a robust lightweight authentication and key establishment protocol that provides higher security levels for <em>IoD</em> environment. The suggested protocol enables authorized users to get real-time information from drones in an <em>IoD</em> environment. It incorporates cost-effective cryptographic operations such as XOR operations and hash functions, as well as fuzzy extractor procedures for user biometric verification. A thorough security analysis demonstrates that our proposed work enhances functioning of security features and is resistant to common attacks such as privileged insider threats, device capture, user impersonation, server impersonation, drone impersonation, and desynchronization attacks. The analytical analysis shows that the proposed protocol minimizes computational and communicational complexities for both the user and the drone when compared to related ones. The proposed protocol is implemented in Java’s Big-Integer class to ensure its appropriateness in a real-time environment.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"282 ","pages":"Pages 851-861"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148574609","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Mitigating Minority Rule: A Game Theoretic Approach to Curbing Voting Bloc Dominance in Blockchain Governance","authors":"Tanuj Surve,&nbsp;Shrikant Tiwari Dr.,&nbsp;Amit Kumar Tyagi Dr.","doi":"10.1016/j.procs.2026.05.080","DOIUrl":"10.1016/j.procs.2026.05.080","url":null,"abstract":"<div><div>Public blockchain networks offer a promising avenue for decentralized decision-making. However, this decentralization is at risk from voting blocs: coordinated minority groups that can disproportionately influence outcomes. This paper delves into the theoretical dynamics of these blocs using game theory and mechanism design. We introduce two innovative bloc resistance mechanisms, inspired by quadratic voting and conviction weighting, to counteract their dominance. While our findings provide a foundation for designing bloc-resistant governance protocols, the challenges of implementation, including unintended consequences and voter collusion, must be carefully navigated. This study underscores the importance of bridging theory with practice to truly harness the democratizing potential of public blockchains.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"282 ","pages":"Pages 326-336"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148531008","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Quantum Inspired Salp Swarm Algorithm (QSSA) Optimized CNN For Chronic Kidney Disease Prediction","authors":"Sukanta Kumar Hati,&nbsp;Nibedan Panda","doi":"10.1016/j.procs.2026.05.082","DOIUrl":"10.1016/j.procs.2026.05.082","url":null,"abstract":"<div><div>Global public health systems are still facing an immense burden of chronic kidney disease (CKD), leading to the development of precise, efficient and interpretable diagnostic methods essential for supporting early detection and intervention. The present research addresses this issue by demonstrating an innovative hybrid deep learning system that brings together a Convolutional Neural Network (CNN) architecture as well as a Quantum-Inspired Salp Swarm Algorithm (QSSA) to enhance predictive performance in the diagnosis of chronic kidney disease. In order to effectively explore the high-dimensional hyperparameter space, the QSSA approach integrates dynamic behavioural modelling of salp swarms using quantum-inspired operators. To improve CNN’s discriminative leadership, key variables are systematically modified, that includes the number of convolutional layers, filter dimensions, learning rate, dropout rate, and batch size. The QSSA-optimized CNN performs better than conventional CNN models optimized with the salp swarm algorithm (SSA) when it comes to of predictive accuracy using medical imaging datasets as CKD. The results highlight the potential of the suggested QSSA-CNN framework as a practical and scalable method in resource-constrained scenarios.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"282 ","pages":"Pages 347-356"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148531010","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Leveraging Large Language Models for Dream Analysis","authors":"Ashish Kumar Srivastava,&nbsp;Md. Asraful Haque,&nbsp;Tauseef Ahmad","doi":"10.1016/j.procs.2026.05.142","DOIUrl":"10.1016/j.procs.2026.05.142","url":null,"abstract":"<div><div>Dream analysis, which has its roots in ancient rituals and psychological theories, now has a modern ally in large language models (LLMs). While traditional approaches frequently rely on subjective interpretation, the introduction of large language models (LLMs) provides a new way to comprehending the intricacies of the dream world. This review paper looks at how LLMs may supplement traditional dream analysis by processing large volumes of dream reports and providing detailed, context-sensitive interpretations. It investigates the substantial advantages of LLMs, including as their capacity to recognize patterns, integrate cross-disciplinary information, and create personalized insights. However, integrating these models presents challenges, such as guaranteeing cultural and symbolic significance, eliminating biases, and preserving interpretative richness without reducing dreams to algorithmic outputs. We assess the current state of development, highlighting recent innovations and case studies where LLMs have been employed for dream research. Finally, we outline future directions, advocating for a collaborative strategy that combines human knowledge with sophisticated artificial intelligence to improve our understanding of dreams and its significant psychological and cultural value.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"282 ","pages":"Pages 1005-1015"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148531285","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hybrid Deep Learning Framework for Real-Time Pest Detection Using YOLOv11 and VGG16","authors":"Srinivas Arukonda,&nbsp;V. Srilakshmi,&nbsp;V. Lakshmi Chetana","doi":"10.1016/j.procs.2026.06.151","DOIUrl":"10.1016/j.procs.2026.06.151","url":null,"abstract":"<div><div>Global agricultural productivity is seriously threatened by pest outbreaks, underscoring the urgent need for intelligent, effective, and interpretable automated detection systems. This study presents a novel hybrid deep learning framework that integrates YOLOv11 for real-time pest localization and VGG16 for enhanced classification accuracy, with Grad-CAM++ providing visual interpretability to increase transparency and user trust. The framework is trained and evaluated on the IP102 dataset, which includes 102 pest categories and poses challenges such as small object detection, class imbalance, and high intra-class visual similarity. The proposed model achieves an Intersection over Union of 90.8% and a mean Average Precision of 98.3%, with precision, recall, and F1-score values of 97.8%, 98.5%, and 98.2%, respectively. Comparative analysis against Faster R-CNN, SSD, EfficientDet, and YOLOv3 confirms its superior accuracy and inference efficiency. Moreover, the framework delivers real-time performance, achieving 142 FPS on an RTX 3090 and 35 FPS on a Jetson Nano, ensuring feasibility on both high-end and edge devices. By combining accurate detection, interpretability, and high-speed processing, this work makes a novel and practical contribution to precision agriculture.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"283 ","pages":"Pages 687-697"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148531348","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Comparative Analysis of Machine Learning Algorithms on Simulated Cyber Attacks Dataset","authors":"Ravi Shankar Jha,&nbsp;Deepika Agarwal,&nbsp;Himanshu Batra,&nbsp;Punit Gupta","doi":"10.1016/j.procs.2026.05.145","DOIUrl":"10.1016/j.procs.2026.05.145","url":null,"abstract":"<div><div>With the rapid expansion of digital connectivity, cybersecurity has become a paramount concern for organizations and individuals alike. The increasing sophistication and diversity of cyber attacks, ranging from Denial-of-Service (DoS) to Man-in-the-Middle (MITM), web exploits, and Malware, poses a critical challenge to traditional security systems. Conventional rule-based detection mechanisms often fail to recognize novel or complex attack patterns, creating a pressing need for automated, intelligent, and data-driven solutions. This research addresses this challenge by employing machine learning (ML) algorithms to analyze and classify network traffic attack classes and normal traffic. The study utilizes a comprehensive, cleaned, balanced, and clustered dataset, ensuring high-quality input for ML models. Baseline algorithms, including Random Forest, Support Vector Machines (SVM), and XGBoost, are applied to perform multiclass classification of cyber attacks. Model performance is evaluated desired matrices evaluation. Additionally, dimensionality reduction and visualization techniques such as PCA and t-SNE, are employed to reveal distinct attack patterns and feature separability within the self-generated dataset. We are also proposing the dataset of diverse attack classes, collected during the experiments, with a demonstration of baseline ML models that can accurately identify diverse cyber threats, highlighting their potential in real-world network security applications. This research emphasizes the importance of a well-preprocessed and balanced dataset in enhancing classification performance, addressing a critical gap where conventional systems often fail. This study mainly focused on two goals: a dataset proposal and a depth analysis of data, special cyber attacks, the implications of it, and a processed guide and impact of it represented with results with benchmark matrices, and their importance.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"282 ","pages":"Pages 1032-1042"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148531287","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Dual-Stage Computational Approach for Liver Tumor Characterization: U-Net 2D Spatial Analysis and XGBoost-Based Classification","authors":"Ritu Gautum,&nbsp;Khurram Izhar,&nbsp;Tarushi Gautam,&nbsp;Chinmaya Sardana,&nbsp;Sahil Rai","doi":"10.1016/j.procs.2026.05.138","DOIUrl":"10.1016/j.procs.2026.05.138","url":null,"abstract":"<div><div>Liver cancer is the sixth most diagnosed cancer globally and the third leading cause of cancer-related deaths, highlighting the need for advanced automated detection systems for early diagnosis. This article presents a dual-stage framework that combines the U-Net 2D deep learning model for segmentation with the XGBoost gradient boosting classifier for data classification. The study utilizes the LiTS2017 benchmark dataset, achieving high average Dice coefficients of 96.84% for liver and 71.35% for tumor segmentation. The XGBoost classifier shows an overall accuracy of 96.78%, with notable precision (97.22%) and recall (95.85%). Compared to other methods, this hybrid system not only localizes tumors effectively but also classifies them accurately while maintaining computational efficiency, processing cases in about 3.15 seconds. The robustness of the approach is confirmed through five-fold cross-validation, indicating a significant step forward in automated liver cancer diagnosis.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"282 ","pages":"Pages 952-966"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148531290","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adaptive Multi-Spectral Remote Sensing Framework for Comprehensive Salt Lake Environmental Monitoring and Groundwater Resource Assessment","authors":"Akey Sungheetha,&nbsp;Rajesh Sharma R,&nbsp;Oluwasegun Julius Aroba","doi":"10.1016/j.procs.2026.05.124","DOIUrl":"10.1016/j.procs.2026.05.124","url":null,"abstract":"<div><div>Accurate and scalable environmental monitoring of ecologically sensitive zones such as salt lake peripheries demands integration of remote sensing physics, adaptive spectral analysis, and deep learning classification. This paper presents a unified adaptive monitoring framework that combines radiometric correction, normalized spectral index computation, deep convolutional neural network (CNN) classification, statistical temporal change detection, and regression-based groundwater level estimation into a coherent operational pipeline. Radiometric correction transforms at-sensor radiance <em>(L_sensor =</em> 45.2 W/m²/sr/µm) to surface reflectance (<em>p surface =</em> 0.247) using atmosphere-geometry parameters <em>(r_atm =</em> 0.82, <em>6</em>_s <em>=</em> 35°, <em>d =</em> 1.012 AU). The Normalized Difference Water Index (NDWI) and Normalized Difference Vegetation Index (NDVI) are computed per pixel to drive both classification and regression. A deep CNN trained on 14-band multispectral imagery achieves 94.7% overall accuracy (Kappa = 0.921) over four land-cover classes—water, vegetation, bare soil, and salt crust—outperforming support vector machines (78.2%) and shallow CNNs (88.1%). Temporal statistical analysis across 72 monthly observations reveals a -23.4% NDVI decline <em>(t_stat =</em> -7.14, <em>p &lt;</em> 0.001) between the reference mean <em>(</em>µ_<em>ref</em> = 0.64) and monitoring mean (µ_<em>mon</em> = 0.49), with an annual degradation rate of -9.4% per year. A calibrated regression model <em>(</em>β₀ <em>=</em> 35.2, β₁ <em>=</em> -28.4, β₂ <em>=</em> -15.7, RMSE = 2.8 m, R² = 0.87) estimates groundwater depth at 12.0 + 2.8 m for a representative pixel. All results are derived through theoretical scientific computation; no real-time devices or operational software were employed. The proposed framework offers a scientifically rigorous and computationally efficient solution for ecosystem degradation monitoring and groundwater assessment.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"282 ","pages":"Pages 796-809"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148531304","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}