PeerJ Computer SciencePub Date : 2025-06-11eCollection Date: 2025-01-01DOI: 10.7717/peerj-cs.2904
Hong Huang, Yang Yang, Yunfei Wang
{"title":"AdvFaceGAN: a face dual-identity impersonation attack method based on generative adversarial networks.","authors":"Hong Huang, Yang Yang, Yunfei Wang","doi":"10.7717/peerj-cs.2904","DOIUrl":"10.7717/peerj-cs.2904","url":null,"abstract":"<p><p>This article aims to reveal security vulnerabilities in current commercial facial recognition systems and promote advancements in facial recognition technology security. Previous research on both digital-domain and physical-domain attacks has lacked consideration of real-world attack scenarios: Digital-domain attacks with good stealthiness often fail to achieve physical implementation, while wearable-based physical-domain attacks typically appear unnatural and cannot evade human visual inspection. We propose AdvFaceGAN, a generative adversarial network (GAN)-based impersonation attack method that generates dual-identity adversarial faces capable of bypassing defenses and being uploaded to facial recognition system databases in our proposed attack scenario, thereby achieving dual-identity impersonation attacks. To enhance visual quality, AdvFaceGAN introduces a structural similarity loss in addition to conventional generative loss and perturbation loss, optimizing the generation pattern of adversarial perturbations. Under the combined effect of these three losses, our method produces adversarial faces with excellent stealthiness that can pass administrator's human review. To improve attack effectiveness, AdvFaceGAN employs an ensemble of facial recognition models with maximum model diversity to calculate identity loss, thereby enhancing similarity to target identities. Innovatively, we incorporate source identity loss into the identity loss calculation, discovering that minor reductions in target identity similarity can be traded for significant improvements in source identity similarity, thus making the adversarial faces generated by our method highly similar to both the source identity and the target identity, addressing limitations in existing impersonation attack methods. Experimental results demonstrate that in black-box attack scenarios, AdvFaceGAN-generated adversarial faces exhibit better stealthiness and stronger transferability compared to existing methods, achieving superior traditional and dual-identity impersonation attack success rates across multiple black-box facial recognition models and three commercial facial recognition application programming interfaces (APIs).</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2904"},"PeriodicalIF":3.5,"publicationDate":"2025-06-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192821/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499184","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
PeerJ Computer SciencePub Date : 2025-06-11eCollection Date: 2025-01-01DOI: 10.7717/peerj-cs.2895
Sun Jae Baek, Minhyeok Lee
{"title":"Recent advances in the inverse design of silicon photonic devices and related platforms using deep generative models.","authors":"Sun Jae Baek, Minhyeok Lee","doi":"10.7717/peerj-cs.2895","DOIUrl":"10.7717/peerj-cs.2895","url":null,"abstract":"<p><p>This article presents an overview of recent research on the inverse design of optical devices using deep generative models. The increasing complexity of modern optical devices necessitates advanced design methodologies that can efficiently navigate vast parameter spaces and generate novel, high-performance structures. Established optimization methods, such as adjoint and topology optimization, have successfully addressed many design challenges. However, the increasing complexity of modern optical devices creates opportunities for complementary approaches. Deep generative models offer additional capabilities by leveraging their ability to learn complex patterns and generate novel designs. This review examines various deep learning methodologies, including multi-layer perceptrons (MLP), convolutional neural networks (CNN), auto-encoders (AE), Generative Adversarial Networks (GAN), and reinforcement learning (RL) approaches. We analyze their applications in the inverse design of photonic devices, comparing their effectiveness and integration in the design process. Our findings indicate that while MLP-based methods were commonly used in early research, recent studies have increasingly employed CNN, GAN, AE, and RL methods, as well as advanced MLP models. Each of these methods offers unique advantages and presents specific challenges in the context of optical device inverse design. This review critically evaluates these deep learning-based inverse design technologies, highlighting their strengths and limitations in the context of optical device design. By synthesizing current research and identifying key trends, this article aims to guide future developments in the application of deep generative models for optical device inverse design.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2895"},"PeriodicalIF":3.5,"publicationDate":"2025-06-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192857/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499310","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"IAESR: IoT-oriented authenticated encryption based on iShadow round function.","authors":"Yanshuo Zhang, Liqiu Li, Hengyu Bao, Xiaohong Qin, Zhiyuan Zhang, Xiaoyi Duan","doi":"10.7717/peerj-cs.2947","DOIUrl":"https://doi.org/10.7717/peerj-cs.2947","url":null,"abstract":"<p><p>With the growing popularity of the Internet of Things (IoT) devices and the widespread application of embedded systems, the demand for security and resource efficiency in these devices is also increasing. Traditional authenticated encryption (AE) algorithms are often unsuitable for lightweight devices due to their complexity and resource consumption, creating a need for lightweight AE algorithms. Lightweight devices typically have limited processing power, storage capacity, and energy resources, which necessitates the design of simple and efficient encryption algorithms that can function within these constraints. Despite these resource limitations, security remains of paramount importance. Therefore, lightweight AE algorithms must minimize resource consumption while ensuring adequate security. This article presents a theoretical lightweight AE scheme based on Shadow, a lightweight block encryption algorithm, to address the requirements for secure communication in resource-constrained environments. The scheme first enhances the Shadow algorithm by introducing the improved Shadow (iShadow) algorithm. It then combines this with the duplex sponge structure to propose the IoT-oriented authenticated encryption based on the iShadow round function (IAESR). The integration of iShadow with the duplex sponge structure achieves a balance between security and efficiency through three key mechanisms: (1) The sponge's capacity (64/128-b for IAESR-32/64) provides provable indistinguishability under chosen-plaintext attack (IND-CPA) and chosen-ciphertext attack (IND-CCA) security bounds, effectively resisting generic attacks with an adversarial advantage limited to <i>O</i>(<i>q</i> <sup>2</sup>/2 <i><sup>c</sup></i> ); (2) the duplex mode's single-pass processing reduces memory overhead by reusing the permutation state; and (3) iShadow's ARX operations reduce energy consumption to 0.4-0.5 µJ/byte on 32-b microcontrollers, outperforming AES-GCM by 20-30%. Empirical tests on an Intel i5-1035G1 CPU demonstrate stable execution times. This design ensures the security and integrity of communication while balancing efficiency, and resource utilization. This design ensures IND-CCA secure confidentiality and integrity against plaintext (INT-PTXT), as demonstrated by the security bounds of the sponge construction. Specifically, IAESR guarantees both confidentiality and authenticity. Additionally, it is particularly well-suited for scenarios with lightweight requirements, such as those found in the IoT.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2947"},"PeriodicalIF":3.5,"publicationDate":"2025-06-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12193455/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499208","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
PeerJ Computer SciencePub Date : 2025-06-11eCollection Date: 2025-01-01DOI: 10.7717/peerj-cs.2543
Malathi Chilakalapudi, Sheela Jayachandran
{"title":"Iterative segmentation and classification for enhanced crop disease diagnosis using optimized hybrid U-Nets model.","authors":"Malathi Chilakalapudi, Sheela Jayachandran","doi":"10.7717/peerj-cs.2543","DOIUrl":"10.7717/peerj-cs.2543","url":null,"abstract":"<p><p>The major challenges that the agricultural sector faces are that with the kind of methodologies that exist, gross limitations may occur to the exact diagnosis of crop diseases. They are unable to achieve correct precision in disease classification, relatively lower accuracy, and delayed response time-all these obstacles result in a deficiency in effectual disease management and control. Our research proposes a new framework instigated and developed to improve crop disease detection and classification by multifaceted analysis. In the core of our methodology is the implementation of adaptive anisotropic diffusion for the denoising of obtained agro images, therefore making it a step towards assurance in data quality. Along with this is the use of a Fuzzy U-Net++ model for image segmentation, whereby fuzzy decisions in generously instill an increase in performance for image segmentation. Feature selection itself is innovated by the introduction of the Moving Gorilla Remora Algorithm (MGRA) combined with convolutional operations, setting a new benchmark in the selection of optimal features pertaining to disease identification operations. To further refine this model, classification is adeptly handled by a process inspired by the LeNet architecture, significantly improving identification against various diseases. Our approach's performance is therefore strongly assessed through a number of renowned datasets, such as PlantVillage and PlantDoc, on which test metrics show superior performance: 8.5% improvement in disease classification precision, 8.3% higher accuracy, 9.4% improved recall, with a reduction in time delay by 4.5%, area under the curve (AUC) increasing by 5.9%, a 6.5% improvement in specificity, far ahead of other methods. This work not only sets new standards in crop disease analysis but also opens possibilities for the preemptive measures to come in agricultural health, promising a future where crop management is more effective and efficient. Our results thus have implications that reach beyond the immediate benefits accruable from improved diagnosis of diseases. It is a harbinger of a new era in agricultural technology where precision, accuracy, and timeliness will meet to enhance crop resilience and yield.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2543"},"PeriodicalIF":3.5,"publicationDate":"2025-06-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12190645/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499249","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Integrating cyber-physical systems with embedding technology for controlling autonomous vehicle driving.","authors":"Manal Abdullah Alohali, Hamed Alqahtani, Abdulbasit Darem, Monir Abdullah, Yunyoung Nam, Mohamed Abouhawwash","doi":"10.7717/peerj-cs.2823","DOIUrl":"10.7717/peerj-cs.2823","url":null,"abstract":"<p><p>Cyber-physical systems (CPSs) in autonomous vehicles must handle highly dynamic and uncertain settings, where unanticipated impediments, shifting traffic conditions, and environmental changes all provide substantial decision-making issues. Deep reinforcement learning (DRL) has emerged as a strong tool for dealing with such uncertainty, yet current DRL models struggle to ensure safety and optimal behaviour in indeterminate settings due to the difficulties of understanding dynamic reward systems. To address these constraints, this study incorporates double deep Q networks (DDQN) to improve the agent's adaptability under uncertain driving conditions. A structured reward system is established to accommodate real-time fluctuations, resulting in safer and more efficient decision-making. The study acknowledges the technological limitations of automobile CPSs and investigates hardware acceleration as a potential remedy in addition to algorithmic enhancements. Because of their post-manufacturing adaptability, parallel processing capabilities, and reconfigurability, field programmable gate arrays (FPGAs) are used to execute reinforcement learning in real-time. Using essential parameters, including collision rate, behaviour similarity, travel distance, speed control, total rewards, and timesteps, the suggested method is thoroughly tested in the TORCS Racing Simulator. The findings show that combining FPGA-based hardware acceleration with DDQN successfully improves computational efficiency and decision-making reliability, tackling significant issues brought on by uncertainty in autonomous driving CPSs. In addition to advancing reinforcement learning applications in CPSs, this work opens up possibilities for future investigations into real-world generalisation, adaptive reward mechanisms, and scalable hardware implementations to further reduce uncertainty in autonomous systems.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2823"},"PeriodicalIF":3.5,"publicationDate":"2025-06-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12190562/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499233","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
PeerJ Computer SciencePub Date : 2025-06-10eCollection Date: 2025-01-01DOI: 10.7717/peerj-cs.2920
Seref Sagiroglu, Ramazan Terzi, Emrah Celtikci, Alp Özgün Börcek, Yilmaz Atay, Bilgehan Arslan, Mustafa Caglar Sahin, Kerem Nernekli, Umut Demirezen, Okan Bilge Ozdemir, Kevser Özdem Karaca, Nuh Azgınoğlu
{"title":"A novel brain tumor magnetic resonance imaging dataset (Gazi Brains 2020): initial benchmark results and comprehensive analysis.","authors":"Seref Sagiroglu, Ramazan Terzi, Emrah Celtikci, Alp Özgün Börcek, Yilmaz Atay, Bilgehan Arslan, Mustafa Caglar Sahin, Kerem Nernekli, Umut Demirezen, Okan Bilge Ozdemir, Kevser Özdem Karaca, Nuh Azgınoğlu","doi":"10.7717/peerj-cs.2920","DOIUrl":"10.7717/peerj-cs.2920","url":null,"abstract":"<p><p>This article presents a new benchmark MRI dataset called the Gazi Brains Dataset 2020, containing MRI images of 100 patients, and introduces initial experimental results performed on this dataset in comparison with available brain MRI datasets. Furthermore, the dataset is analyzed using eight different deep learning models for high-grade glioma tumor prediction, classification, and detection tasks. Additionally, this study demonstrates the results of an explainable Artificial Intelligence (XAI) approach applied to the trained models. To demonstrate the utility of the proposed dataset, different deep learning models were applied to the problem, and these models were tested on various data and models applied for various tasks such as region of interest extraction, whole tumor segmentation, prediction, detection, and classification with accuracy, precision, recall, and F1-score. The experimental results indicate that the dataset is highly effective for multiple purposes, and the models reached significant results with successful F1-scores ranging between 93.2% and 96.4%. ROI and whole tumor segmentations were successfully performed and compared with seven algorithms with accuracies of 87.61% and 97.18%. The Grad-CAM model also demonstrated satisfactory accuracy across the tests that were conducted. Moreover, this study explores the application of XAI to the trained models, providing interpretability and insights into the decision-making processes. The findings signify that this dataset holds significant potential for various future research directions, including age estimation, gender detection, causal inference with XAI, and disease-related survival analysis.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2920"},"PeriodicalIF":3.5,"publicationDate":"2025-06-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192951/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499146","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
PeerJ Computer SciencePub Date : 2025-06-09eCollection Date: 2025-01-01DOI: 10.7717/peerj-cs.2902
Manzoor Ahmed, Touseef Hussain, Muhammad Shahwar, Feroz Khan, Muhammad Sheraz, Wali Ullah Khan, Teong Chee Chuah, It Ee Lee
{"title":"Intelligent reflecting surface backscatter-enabled physical layer security enhancement <i>via</i> deep reinforcement learning.","authors":"Manzoor Ahmed, Touseef Hussain, Muhammad Shahwar, Feroz Khan, Muhammad Sheraz, Wali Ullah Khan, Teong Chee Chuah, It Ee Lee","doi":"10.7717/peerj-cs.2902","DOIUrl":"10.7717/peerj-cs.2902","url":null,"abstract":"<p><p>This article introduces a novel strategy for wireless communication security utilizing intelligent reflecting surfaces (IRS). The IRS is strategically deployed to mitigate jamming attacks and eavesdropper threats while improving signal reception for legitimate users (LUs) by redirecting jamming signals toward desired communication signals leveraging physical layer security (PLS). By integrating the IRS into the backscatter communication system, we enhance the overall secrecy rate of LU, by dynamically adjusting IRS reflection coefficients and active beamforming at the base station (BS). A design problem is formulated to jointly optimize IRS reflecting beamforming and BS active beamforming, considering time-varying channel conditions and desired secrecy rate requirements. We propose a novel approach based on deep reinforcement learning (DRL) named Deep-PLS. This approach aims to determine an optimal beamforming policy capable of thwarting eavesdroppers in evolving environmental conditions. Extensive simulation studies validate the efficacy of our proposed strategy, demonstrating superior performance compared to traditional IRS approaches, IRS backscattering-based anti-eavesdropping methods, and other benchmark strategies in terms of secrecy performance.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2902"},"PeriodicalIF":3.5,"publicationDate":"2025-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192987/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499234","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Automatic dependent surveillance-broadcast (ADS-B) anomalous messages and attack type detection: deep learning-based architecture.","authors":"Waqas Ahmed, Ammar Masood, Jawad Manzoor, Sedat Akleylek","doi":"10.7717/peerj-cs.2886","DOIUrl":"10.7717/peerj-cs.2886","url":null,"abstract":"<p><p>Automatic Dependent Surveillance-Broadcast (ADS-B) is a vital communication protocol within air traffic control (ATC) systems. Unlike traditional technologies, ADS-B utilizes the Global Positioning System (GPS) to deliver more accurate and precise location data while reducing operational and deployment costs. It enhances radar coverage and serves as a standalone solution in areas lacking radar services. Despite these advantages, ADS-B faces significant security vulnerabilities due to its open design and the absence of built-in security features. Given its critical role, developing an advanced security framework to classify ADS-B messages and identify various attack types is essential to safeguard the system. This research makes several key contributions to address these challenges. First, it presents a comprehensive review of state-of-the-art machine learning and deep learning techniques, critically analyzing existing methodologies for ADS-B intrusion detection. Second, a detailed attack model is developed, categorizing potential threats and aligning them with key security requirements, including confidentiality, integrity, availability, and authentication. Third, the study proposes a robust and accurate Intrusion Detection System (IDS) using three advanced deep learning models-TabNet, Neural Oblivious Decision Ensembles (NODE), and DeepGBM-to classify ADS-B messages and detect specific attack types. The models are evaluated using standard metrics, including accuracy, precision, recall, and F1-score. Among the tested models, DeepGBM achieves the highest accuracy at 98%, outperforming TabNet (92%) and NODE (96%). The findings offer valuable insights into ADS-B security and define essential requirements for a future security framework, contributing actionable recommendations for mitigating threats in this critical communication protocol.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2886"},"PeriodicalIF":3.5,"publicationDate":"2025-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192918/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499216","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
PeerJ Computer SciencePub Date : 2025-06-05eCollection Date: 2025-01-01DOI: 10.7717/peerj-cs.2934
Xinting Yang, Zehua Zhang
{"title":"Enhancing book genre classification with BERT and InceptionV3: a deep learning approach for libraries.","authors":"Xinting Yang, Zehua Zhang","doi":"10.7717/peerj-cs.2934","DOIUrl":"10.7717/peerj-cs.2934","url":null,"abstract":"<p><p>Accurate book genre classification is essential for library organization, information retrieval, and personalized recommendations. Traditional classification methods, often reliant on manual categorization and metadata-based approaches, struggle with the complexities of hybrid genres and evolving literary trends. To address these limitations, this study proposes a hybrid deep learning model that integrates visual and textual features for enhanced genre classification. Specifically, we employ InceptionV3, an advanced convolutional neural network architecture, to extract visual features from book cover images and bidirectional encoder representations from transformers (BERT) to analyze textual data from book titles. A scaled dot-product attention mechanism is used to effectively fuse these multimodal features, dynamically weighting their contributions based on contextual relevance. Experimental results on the BookCover30 dataset demonstrate that our proposed model outperforms baseline approaches, achieving a balanced accuracy of 0.7951 and an F1-score of 0.7920, surpassing both standalone image- and text-based classifiers. This study highlights the potential of deep learning in improving automated genre classification, offering a scalable and adaptable solution for libraries and digital platforms. Future research may focus on expanding dataset diversity, optimizing computational efficiency, and addressing biases in classification models.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2934"},"PeriodicalIF":3.5,"publicationDate":"2025-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12193415/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499269","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
PeerJ Computer SciencePub Date : 2025-06-05eCollection Date: 2025-01-01DOI: 10.7717/peerj-cs.2883
Fengjun Qi, Zhenping Liu, Wenzheng Zhang, Zhenjie Sun
{"title":"Optimization of multi-objective feature regression models for designing performance assessment methods in college and university educational reform.","authors":"Fengjun Qi, Zhenping Liu, Wenzheng Zhang, Zhenjie Sun","doi":"10.7717/peerj-cs.2883","DOIUrl":"10.7717/peerj-cs.2883","url":null,"abstract":"<p><p>The evaluation of teacher performance in higher education is a critical component of educational reform, requiring robust and accurate assessment methodologies. Multi-objective regression offers a promising approach to optimizing the construction of performance evaluation index systems. However, conventional regression models often rely on a shared input space for all targets, neglecting the fact that distinct and complex feature sets may influence each target. This study introduces a novel Multi-Objective Feature Regression model under Label-Specific Features (MOFR-LSF), which integrates target-specific features and inter-target correlations to address this limitation. By extending the single-objective stacking framework, the proposed method learns label-specific features for each target and employs cluster analysis on binned samples to uncover underlying correlations among objectives. Experimental evaluations on three datasets-Education Reform (EDU-REFORM), Programme for International Student Assessment (PISA), and National Assessment of Educational Progress (NAEP)-demonstrate the superior performance of MOFR-LSF, achieving relative root mean square error (RRMSE) values of 0.634, 0.332, and 0.925, respectively, outperforming existing multi-objective regression algorithms. The proposed model not only enhances predictive accuracy but also strengthens the scientific validity and fairness of performance evaluations, offering meaningful contributions to educational reform in colleges and universities. Moreover, its adaptable framework suggests potential applicability across a range of other domains.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2883"},"PeriodicalIF":3.5,"publicationDate":"2025-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192971/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499300","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}