PeerJ Computer Science最新文献

{"title":"Comparative performance of twelve machine learning models in predicting COVID-19 mortality risk in children: a population-based retrospective cohort study in Brazil.","authors":"Adriano Lages Dos Santos, Maria Christina L Oliveira, Enrico A Colosimo, Robert H Mak, Clara C Pinhati, Stella C Gallante, Hercílio Martelli-Júnior, Ana Cristina Simões E Silva, Eduardo A Oliveira","doi":"10.7717/peerj-cs.2916","DOIUrl":"10.7717/peerj-cs.2916","url":null,"abstract":"<p><p>The COVID-19 pandemic has catalyzed the application of advanced digital technologies such as artificial intelligence (AI) to predict mortality in adult patients. However, the development of machine learning (ML) models for predicting outcomes in children and adolescents with COVID-19 remains limited. This study aimed to evaluate the performance of multiple machine learning models in forecasting mortality among hospitalized pediatric COVID-19 patients. In this cohort study, we used the SIVEP-Gripe dataset, a public resource maintained by the Ministry of Health, to track severe acute respiratory syndrome (SARS) in Brazil. To create subsets for training and testing the machine learning (ML) models, we divided the primary dataset into three parts. Using these subsets, we developed and trained 12 ML algorithms to predict the outcomes. We assessed the performance of these models using various metrics such as accuracy, precision, sensitivity, recall, and area under the receiver operating characteristic curve (AUC). Among the 37 variables examined, 24 were found to be potential indicators of mortality, as determined by the chi-square test of independence. The Logistic Regression (LR) algorithm achieved the highest performance, with an accuracy of 92.5% and an AUC of 80.1%, on the optimized dataset. Gradient boosting classifier (GBC) and AdaBoost (ADA), closely followed the LR algorithm, producing similar results. Our study also revealed that baseline reduced oxygen saturation, presence of comorbidities, and older age were the most relevant factors in predicting mortality in children and adolescents hospitalized with SARS-CoV-2 infection. The use of ML models can be an asset in making clinical decisions and implementing evidence-based patient management strategies, which can enhance patient outcomes and overall quality of medical care. LR, GBC, and ADA models have demonstrated efficiency in accurately predicting mortality in COVID-19 pediatric patients.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2916"},"PeriodicalIF":3.5,"publicationDate":"2025-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192853/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499228","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":"RBI: a novel algorithm for regulatory-metabolic network model in designing the optimal mutant strain.","authors":"Ridho Ananda, Kauthar Mohd Daud, Suhaila Zainudin","doi":"10.7717/peerj-cs.2880","DOIUrl":"10.7717/peerj-cs.2880","url":null,"abstract":"<p><p>Over the last 20 years, researchers have proposed regulatory-metabolic network models to integrate gene regulatory networks (GRNs) and metabolic networks in <i>in silico</i> metabolic engineering, aiming to enhance the production rate of desired metabolites. However, the proposed models are unable to comprehensively include the Boolean rules in the empirical gene regulatory networks (GRNs) and gene-protein-reaction (GPR) interactions. Thus, the types of gene interactions, such as inhibition and activation, are disregarded from the analysis. This may result in sub-optimal model performance. Hence, this article presented a novel model using reliability theory to include Boolean rules in empirical GRNs and GPR rules in the integrating process. The proposed algorithm of this model is termed as a reliability-based integrating (RBI) algorithm. The suggested algorithm had three variants: RBI-T1, RBI-T2, and RBI-T3. The performance of the RBI algorithms was assessed by comparing them with the existing algorithms, using empirical results and validated transcription factors (TF) knockout schemes, and their complexity time was identified. Also, the RBI method was implemented in the design of optimal mutant strains of <i>Escherichia coli</i> and <i>Saccharomyces cerevisiae</i>. The simulation results indicated that the effectiveness and efficiency of the RBI algorithms are adequately strong and competitive relative to the existing algorithms. Furthermore, the RBI algorithm effectively identified eight schemes capable of enhancing succinate and ethanol production rates by maintaining the survival of microbial strains. Those results demonstrated that the RBI algorithms are recommended for the construction of optimum mutant strains in <i>in silico</i> metabolic engineering.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2880"},"PeriodicalIF":3.5,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12199197/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144509429","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":"Optimized deep learning approach for lung cancer detection using flying fox optimization and bidirectional generative adversarial networks.","authors":"Manal Abdullah Alohali, Hamed Alqahtani, Shouki A Ebad, Faiz Abdullah Alotaibi, Venkatachalam K, Jaehyuk Cho","doi":"10.7717/peerj-cs.2853","DOIUrl":"10.7717/peerj-cs.2853","url":null,"abstract":"<p><p>Lung cancer remains one of the most prevalent and life-threatening diseases, often diagnosed at an advanced stage due to the challenges in early detection. Contributory factors include genetic mutations, smoking, alcohol consumption, and exposure to hazardous environmental conditions. Computer-aided diagnosis (CAD) systems have significantly improved early cancer detection, but limitations such as high-dimensional feature sets and overfitting issues persist. This study presents an optimised deep learning approach for lung cancer classification, integrating flying fox optimization (FFXO) for feature selection and bidirectional generative adversarial networks (Bi-GAN) for classification. The methodology consists of three key phases: (1) Data preprocessing, where missing values are handled using the multiple imputations by chain equation (MICE) technique and feature scaling is applied using standard and min-max scalers; (2) Feature selection, where the FFXO algorithm reduces feature dimensionality to enhance classification efficiency; and (3) Lung tumor classification, utilizing Bi-GAN to improve predictive accuracy. The proposed system was evaluated using key performance metrics-accuracy, precision, recall, and F1-score-and demonstrated superior performance to conventional models. Experimental results on a publicly available lung cancer dataset showed an accuracy of 98.7% highlighting the approach's robustness in precise lung tumor classification. This study provides a novel framework for improving the reliability and efficiency of lung cancer detection, offering significant potential for clinical applications.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2853"},"PeriodicalIF":3.5,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192728/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499302","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":"Yoga pose recognition using dual structure convolutional neural network.","authors":"Xiang Meng, Zhaobing Liu","doi":"10.7717/peerj-cs.2907","DOIUrl":"10.7717/peerj-cs.2907","url":null,"abstract":"<p><p>As a popular form of physical and mental exercise, the correct execution of yoga movements is crucial. With the development of deep learning technologies, automatic recognition of yoga postures has become popular. To recognize five different yoga postures, this article proposed a dual structure convolutional neural network with a feature fusion function, which consists of the convolutional neural network A (CNN A) and convolutional neural network B (CNN B). Among them, the structure CNN A observes different channels finding the global feature of yoga images, and the structure CNN B calculates the depth information in each pixel of the yoga images. Following that, the extracted global feature and local feature are fused by a feature fusion function of taking a matrix dot multiplication. Finally, the softmax layer accurately recognizes yoga postures based on the fused features. Experimental results show that the proposed model achieves 97.23% accuracy with 96.08% precision and defeats against the competitors in the recognition of yoga postures. Moreover, the feature fusion function is proved to be successful in terms of the recognition to yoga postures. We also find that the feature fusion with a matrix dot multiplication operation can significantly improve the recognition accuracy of yoga postures than that with a direct connection operation.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2907"},"PeriodicalIF":3.5,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192719/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499319","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":"Efficient deep learning model for classifying lung cancer images using normalized stain agnostic feature method and FastAI-2.","authors":"Pranshu Saxena, Sanjay Kumar Singh, Mamoon Rashid, Sultan S Alshamrani, Mrim M Alnfiai","doi":"10.7717/peerj-cs.2903","DOIUrl":"10.7717/peerj-cs.2903","url":null,"abstract":"<p><strong>Background: </strong>Lung cancer has the highest global fatality rate, with diagnosis primarily relying on histological tissue sample analysis. Accurate classification is critical for treatment planning and patient outcomes.</p><p><strong>Methods: </strong>This study develops a computer-assisted diagnosis system for non-small cell lung cancer histology classification, utilizing the FastAI-2 framework with a modified ResNet-34 architecture. The methodology includes stain normalization using LAB colour space for colour consistency, followed by deep learning-based classification. The proposed model is trained on the LC25000 dataset and compared with VGG11 and SqueezeNet1_1, demonstrating modified ResNet-34's optimal balance between depth and performance. FastAI-2 enhances computational efficiency, enabling rapid convergence with minimal training time.</p><p><strong>Results: </strong>The proposed system achieved 99.78% accuracy, confirming the effectiveness of automated lung cancer histopathology classification. This study highlights the potential of artificial intelligence (AI)-driven diagnostic tools to assist pathologists by improving accuracy, reducing workload, and enhancing decision-making in clinical settings.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2903"},"PeriodicalIF":3.5,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192963/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499264","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":"A collaborative data storage with incentive mechanism for blockchain-based IoV.","authors":"Quan Shi, Lankai Wang, Chen Chen","doi":"10.7717/peerj-cs.2849","DOIUrl":"10.7717/peerj-cs.2849","url":null,"abstract":"<p><p>As the volume of data in the Internet of Vehicles (IoV) continues to grow, challenges such as insufficient storage capacity and potential privacy breaches become more pronounced. To address these issues, this article proposes a novel collaborative data storage scheme with an incentivization mechanism, termed Blockchain-Based Collaborative Data Storage with Incentive Mechanism for IoV (CDS-BIoV). The CDS-BIoV framework consists of vehicles, roadside units (RSUs), and cloud infrastructure. In the first phase, vehicles collect and transmit data to their nearest RSU nodes. To encourage active participation in data reception and storage, an incentive mechanism is introduced to motivate RSU nodes. Two algorithms are developed: the Incentive Mechanism Collaborative Data Storage Algorithm (I-CDSA) and the Data Offloading Algorithm (DOA). The I-CDSA uses a competitiveness matrix to incentivize RSU nodes to minimize storage consumption, while the DOA employs incentives to secure additional cloud storage for offloading data. Experimental results show that the CDS-BIoV scheme reduces storage consumption by up to 93% compared to the Generic Parallel Database (GPDB), particularly as the number of blocks increases, effectively alleviating storage capacity limitations.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2849"},"PeriodicalIF":3.5,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192716/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499137","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":"Application of domain-specific modeling in kinetography and bipedal humanoid robot control.","authors":"Verislav Djukić, Dragana Oros, Marko Penčić, Zhenli Lu","doi":"10.7717/peerj-cs.2864","DOIUrl":"10.7717/peerj-cs.2864","url":null,"abstract":"<p><p>The article presents a new approach in the development of software for bipedal humanoid robot controllers, based on the construction and application of graphic domain-specific languages (DSLs). The notations used to describe dance movements and gestures are typical examples of DSLs. With certain extensions, related to the description of foot topology, sensors and actuators, such DSLs are applicable for modeling dance movements that would be performed by a robot. The existing software development methodologies in robotics have a purely mechanistic approach to understanding and implementing robotic tasks. Such an approach in humanoid robotics complicates the understanding of the problem, as well as the specification and implementation of solutions. Our approach, which uses DSLs, adopts complex movements and gestures performed by the feet of dancers using professional dancers, people with above-average motor skills, as reference. We believe that the developed software can also be successfully applied to assistive robots that would help people with special needs whose mobility is significantly lower than average.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2864"},"PeriodicalIF":3.5,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192693/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499214","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":"Class-weighted Dempster-Shafer in dual-level fusion for multimodal fake real estate listings detection.","authors":"Maifuza Mohd Amin, Nor Samsiah Sani, Mohammad Faidzul Nasrudin","doi":"10.7717/peerj-cs.2797","DOIUrl":"10.7717/peerj-cs.2797","url":null,"abstract":"<p><strong>Background: </strong>Detecting fake multimodal property listings is a significant challenge in online real estate platforms due to the increasing sophistication of fraudulent activities. The existing multimodal data fusion methods have several limitations and strengths in identifying fraudulent listings. Single-level fusion models whether at the feature, decision, or intermediate level struggle with balancing the contributions of different modalities leading to suboptimal decision-making. To address these problems, a dual-level fusion from multimodal for fake real estate listings detection is proposed. The dual-level fusion allows the integration of detailed features from text and image data to be performed at an early stage, followed by the metadata fusion at the decision stage in order to obtain a more comprehensive final classification. Furthermore, a new weighting scheme is introduced to optimize Dempster-Shafer in decision fusion to help the model achieve optimal performance and as a result, our method improves the classification. The Dempster-Shafer without class weightage lacks the flexibility to adapt to varying levels of uncertainty or importance across different classes.</p><p><strong>Methods: </strong>In Class Weighted Dempster-Shafer in Dual Level Fusion (CWDS-DLF), we employ advanced models (XLNet for text and ResNet101 for images) for feature extraction and use the Dempster-Shafer theory for decision fusion. A new weighting scheme, based on Bayesian optimization, was used to assign optimal weights to the 'fake' and 'not fake' classes, thereby enhancing the Dempster-Shafer theory in the decision fusion process.</p><p><strong>Results: </strong>The CWDS-DLF was evaluated on the property listing website dataset and achieved an F1 score of 96% and an accuracy of 93%. A t-test confirms the significance of these improvements (<i>p</i> < 0.05), demonstrating the effectiveness of our method in detecting fake property listings. Compared to other models, including 2D-convolutional neural network (CNN), XGBoost, and various multimodal approaches, our model consistently outperforms in precision, recall, and F1-score. This underscores the potential of integrating multimodal analysis with sophisticated fusion techniques to enhance the detection of fake property listings, ultimately improving consumer protection and operational efficiency in online real estate platforms.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2797"},"PeriodicalIF":3.5,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12190670/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499224","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":"High-fidelity steganography in EEG signals using advanced transform-based methods.","authors":"Enes Efe","doi":"10.7717/peerj-cs.2900","DOIUrl":"10.7717/peerj-cs.2900","url":null,"abstract":"<p><p>The increasing prevalence of digital health solutions and smart health devices (SHDs) ensures the continuity of personal biometric data while simultaneously raising concerns about their security and privacy. Consequently, the development of novel encryption techniques and data protection policies is crucial to comply with regulations such as The Health Insurance Portability and Accountability Act (HIPAA) and to safeguard against cyber threats. This study introduces a robust and efficient method for embedding private information into electroencephalogram (EEG) signals by employing the stationary wavelet transform (SWT), singular value decomposition (SVD), and tent map techniques. The proposed approach aims to increase embedding capacity while maintaining signal integrity, ensuring resilience against various forms of distortion, and achieving computational efficiency. Experiments were conducted on three publicly available EEG datasets (Graz A, DEAP, and Bonn), and performance was evaluated using widely recognized metrics, including peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), percentage root mean square difference (PRD), normalized cross-correlation (NCC), bit error rate (BER), and Euclidean distance (ED). The results indicate that the method preserves perceptual quality, achieving PSNR values above 60 dB and demonstrating minimal signal distortion. Robustness tests involving noise addition, random cropping, and low-pass filtering confirm the method's high resilience, with BER approaching zero and NCC near unity. Moreover, the proposed method demonstrates significantly reduced hiding and extraction times compared to conventional approaches, enhancing its suitability for real-time, secure biomedical data transmission.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2900"},"PeriodicalIF":3.5,"publicationDate":"2025-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192647/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499204","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":"Enhancing east-west interface security in heterogeneous SDN <i>via</i> blockchain.","authors":"Hamad Alrashede, Fathy Eassa, Abdullah Marish Ali, Hosam Aljihani, Faisal Albalwy","doi":"10.7717/peerj-cs.2914","DOIUrl":"https://doi.org/10.7717/peerj-cs.2914","url":null,"abstract":"<p><p>Software defined networking (SDN) increasingly integrates multiple controllers from diverse vendors to enhance network scalability, flexibility, and reliability. However, such heterogeneous deployments pose significant security threats, especially at the east-west interface which is connecting these controllers. Existing solutions are inadequate for ensuring robust protection across multi-vendor SDN environments as most of them are meant to a specific type of attacks, use centralized solution, or designed for homogeneous SDN environments. This study proposes a blockchain-based security framework to address existing security gaps within heterogeneous SDN environments. The framework establishes a decentralized, robust, and interoperable security layer for distributed SDN controllers. By utilizing the Ethereum blockchain with customized smart contract-based checks, the proposed approach enables mutual authentication among controllers, secures data exchange, and controls network access. The framework effectively mitigates common SDN threats such as distributed denial-of-service (DDoS), man-in-the-middle (MitM), false data injection, and unauthorized access. Experimental results highlight the practicality of the solution, achieving a stable throughput of approximately 20 transactions per second with an average authentication latency of 28-40 ms. These results demonstrate that the proposed framework not only enhances inter-controller communication security but also maintains the network performance, making it a reliable and scalable solution for real-world SDN deployments.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2914"},"PeriodicalIF":3.5,"publicationDate":"2025-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12192836/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499272","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}