PeerJ Computer Science最新文献

The rapid growth in wireless communication demands has led to a surge in research on technologies capable of enhancing communication reliability, coverage, and energy efficiency. Among these, uncrewed aerial vehicles (UAV) and reconfigurable intelligent surfaces (RIS) have emerged as promising solutions. Prior research on using deep reinforcement learning (DRL) to integrate RIS with UAV concentrated on enhancing signal quality and coverage, but it ignored the challenges caused by electromagnetic interference (EMI). This article introduces a novel framework addressing the challenges posed by EMI from Gallium nitride (GaN) power amplifiers in RIS-assisted UAV communication systems. By integrating DRL with quadrature phase shift keying (QPSK) modulation, the proposed system dynamically optimizes UAV deployment and RIS configurations in real-time, mitigating EMI effects, improving signal-to-interference-plus-noise ratio (SINR), and enhancing energy efficiency. The framework demonstrates superior performance, with an SINR improvement of up to 6.5 dB in interference-prone environments, while achieving a 38% increase in energy efficiency compared to baseline models. Additionally, the system significantly reduces EMI impact, with a mitigation rate of over 70%, and extends coverage area by 35%. The integration of QPSK and DRL allows for real-time decision-making that balances communication quality and energy consumption. These results show the system's potential to outperform traditional methods, particularly in dynamic and challenging environments such as urban, disaster recovery, and remote settings.

Background: The rapid development of artificial intelligence has permeated many fields, with its application in drug discovery becoming increasingly mature. Machine learning, particularly deep learning, has significantly improved the efficiency of drug discovery. In the core task of predicting drug-target affinity (DTA), deep learning enhances predictive performance by automatically extracting complex features from compounds and proteins.

Methods: Traditional approaches often rely heavily on sequence and two-dimensional structural information, overlooking critical three-dimensional and physicochemical properties. To address this, we propose a novel model-Cross Attention Fusion based on Information Enhancement for Drug-Target Affinity Prediction (CAFIE-DTA)-which incorporates protein 3D curvature and electrostatic potential information. The model approximates protein surface curvature using Delaunay triangulation, calculates total electrostatic potential via Adaptive Poisson-Boltzmann Solver (APBS) software, and employs cross multi-head attention to fuse physicochemical and sequence information of proteins. Simultaneously, it integrates graph-based and physicochemical features of compounds using the same attention mechanism. The resulting protein and compound vectors are concatenated for affinity prediction.

Results: Cross-validation and comparative evaluations on the benchmark Davis and KIBA datasets demonstrate that CAFIE-DTA outperforms existing methods. On the Davis dataset, it achieved improvements of 0.003 in confidence interval (CI) and 0.022 in R². On the KIBA dataset, it improved MSE by 0.008, CI by 0.005, and R² by 0.017. Compared to traditional models relying on 2D structures and sequence data, CAFIE-DTA shows superior performance in DTA prediction. The source code is available at: https://github.com/NTU-MedAI/CAFIE-DTA.

Background: This study presents a new approach to decision-making based on the selection of decision-makers according to evaluated criteria in multi-criteria decision-making (MCDM) methods. Therefore, sub-decision-maker groups (SDMGs) are created for each evaluated criterion. The SDMG approach, which is created according to the criteria, offers a more flexible and dynamic structure than the existing approaches. This approach aims to use the expertise and knowledge of decision-makers more effectively. The decision-making approach presented in this study offers an innovative model and adds a new dimension to decision-making processes. This decision-making approach is applied to the plastic injection moulding machine risk assessment, as it involves different criteria. In addition to classical risk parameters such as probability, severity, frequency, and detectability, new parameters such as human error, machine error, and existing safety measures are also used in the risk assessment.

Methods: The integration of the analytic hierarchy process (AHP) and the technique for order preference by similarity to ideal solution (TOPSIS) methods into the interval valued fermatean fuzzy set (IVFFS) environment makes an important contribution to a more comprehensive consideration of risks and uncertainties in the risk assessment process. The IVFF-AHP method is used to weight the risk parameters and determine the hazard scores, and the TOPSIS method is used to rank the hazards. A holistic and systematic approach to risk assessment has been achieved by integrating these two methods. Modelling of these methods is carried out using MATLAB_R2024a software.

Results: According to the evaluated criteria, it was concluded that the determination of the decision makers separately is applicable to the decision-making process. Identifying the existing safety measures parameter as the most important risk parameter emphasizes the central role of this factor in risk assessment. In addition, machine error and human error parameters are also found to be important in risk assessment. These parameters, which are used for the first time in the literature, offer a broader perspective than traditional methods and provide significant advantages in risk assessment. According to the evaluations, electricity, asphyxiating and toxic gases, and hot water use are determined as the most risky hazards. The sensitivity and comparative analysis performed in the study confirm that the proposed methodology produces consistent and reasonable results.

Virtual reality (VR) holds significant promise for enhancing science education by providing immersive and interactive learning experiences. However, the optimal interaction modality within educational VR environments remains an open question. This study investigates the impact of hand-based vs. controller-based interaction on sixth-grade students' sense of presence and interaction performance in a VR science laboratory simulation. Fifty-four sixth-grade students were randomly assigned to either a hand-based interaction group or a controller-based interaction group. Participants completed three interactive science experiments (solar system, electrical circuits, and force/energy) within a virtual laboratory environment designed to mimic their school's physical lab. Presence was assessed using a validated Turkish adaptation of the Presence Questionnaire (PQ), while interaction performance was evaluated using a structured observation form completed by a school teacher. Independent samples t-tests and Mann-Whitney U tests were used to compare the presence and performance scores between the groups. Supplementary analyses explored the effects of gender and prior VR experience. Contrary to expectations, no significant differences were found in either presence (t(49.4) = -0.01, p = 0.992) or interaction performance (t(52) = -1.30, p = 0.199) between the hand-based and controller-based interaction groups. Both interaction modalities yielded comparable levels of self-reported presence and observed performance. However, an unexpected finding emerged regarding performance. A supplementary analysis revealed a significant main effect of gender on performance scores (F(1, 50) = 4.844, p = 0.032), independent of interaction type. Specifically, males demonstrated significantly higher performance than females. This study suggests that, for sixth-grade students engaging in these specific VR science simulations, hand-based and controller-based interactions are equally effective in terms of fostering presence and supporting interaction performance. These findings have practical implications for the design and implementation of VR learning environments, particularly in resource-constrained settings where the reduced maintenance and hygiene concerns associated with hand-based interaction may be advantageous.

Early diagnosis of breast cancer through periodic screening is a vital ally in the fight for survival. Mammography, recognized as one of the most widely used and cost-effective tools for detecting early signs of asymmetry, calcification, masses, and architectural distortion in breast tissue, plays a significant role in nearly all screening scenarios. However, the interpretation and scoring of mammograms is a complex multi-parameter process that frequently leads to false-positive and false-negative results. This article introduces a new deep-learning-based model that classifies mammograms according to the Breast Imaging Reporting and Data System (BI-RADS) assessment categories. The model is trained on a private dataset, intentionally excluding no BI-RADS categories. A novel deep neural network architecture is employed to more accurately classify breasts, including their boundaries, as regions of interest (ROIs). The ConvNeXt architecture serves as a feature extractor for lower-level features, which are then combined with the layers of a randomly initialized naive inception module to capture higher-level features. Diagnosis is achieved through three experimental tests, yielding accuracy rates ranging from 82.08% to 86.27%. These promising accuracy levels, in comparison to previous studies, can be attributed to a more comprehensive approach to addressing BI-RADS scoring challenges. In addition to pursuing further enhancements in accuracy, future research should consider integrating prior radiology reports to create a more realistic end-to-end computer-aided detection system.

Course and instructor evaluations (CIE) are essential tools for assessing educational quality in higher education. However, traditional CIE systems often suffer from inconsistencies between structured responses and open-ended feedback, leading to unreliable insights and increased administrative workload. This study suggests a modular system to address these challenges, leveraging sentiment analysis and inconsistency detection to enhance the reliability and efficiency of CIE processes.

Background: Improving the reliability of CIE data is crucial for informed decision-making in higher education. Existing methods fail to address discrepancies between numerical scores and textual feedback, resulting in misleading evaluations. This study proposes a system to identify and exclude inconsistent data, providing more reliable insights.

Methods: Using the Design Science Research Methodology (DSRM), a system architecture was developed with five modules: data collection, preprocessing, sentiment analysis, inconsistency detection, and reporting. A dataset of 13,651 anonymized Turkish CIE records was used to train and evaluate machine learning algorithms, including support vector machines, naive Bayes, random forest, decision trees, K-nearest neighbors, and OpenAI's GPT-4 Turbo Preview model. Sentiment analysis results from open-ended responses were compared with structured responses to identify inconsistencies.

Results: The GPT-4 Turbo Preview model outperformed traditional algorithms, achieving 85% accuracy, 88% precision, and 95% recall. Analysis of a prototype system applied to 431 CIEs identified a 37% inconsistency rate. By excluding inconsistent data, the system generated reliable reports with actionable insights for course and instructor performance. The purpose of this study is to design and evaluate a new system using the Design Science Research (DSR) approach to enhance the accuracy and reliability of course evaluation processes employed in higher education institutions. The modular system effectively addresses inconsistencies in CIE processes, offering a scalable and adaptable solution for higher education institutions. By integrating advanced machine learning techniques, the system enhances the accuracy and reliability of evaluation reports, supporting data-driven decision-making. Future work will focus on refining sentiment analysis for neutral comments and broadening the system's applicability to diverse educational contexts. This innovative approach represents a significant advancement in leveraging technology to improve educational quality.

Next generation sequencing (NGS) has become a cornerstone of modern genomics, enabling high-throughput analysis of DNA and RNA with wide applications across medicine, research, and biotechnology. However, the growing adoption of NGS technologies has introduced significant cyber-biosecurity risks, particularly those arising from insider threats and organizational shortcomings. While technical vulnerabilities have received attention, the human and behavioral dimensions of cybersecurity in NGS environments remain underexplored. This study investigates the role of human factors and organizational behavior in shaping cyber-biosecurity risks in NGS workflows. A mixed-method approach was employed, combining survey data from 120 participants across four countries with statistical analyses including chi-square tests, cross-tabulations, and cluster analysis. The study assessed cybersecurity training availability, employee engagement, training effectiveness, and awareness of insider threats. Findings reveal substantial gaps in training frequency and participation, with 36% of respondents reporting no access to NGS-specific cybersecurity training. Only a minority of participants felt confident in detecting cyber threats, and 32.5% had never applied cybersecurity knowledge in practice. Chi-square results indicate significant associations between training frequency and threat recognition, training relevance, and knowledge application. Cluster analysis further categorized organizations into "robust," "moderate," and "emergent" cybersecurity maturity profiles. The study offers an evidence-based framework to enhance cyber-biosecurity in NGS settings by addressing human-centric risks. It recommends role-specific training, frequent policy updates, and improved organizational communication to mitigate insider threats. These insights support the development of targeted interventions and policies to strengthen the cybersecurity culture in genomics organizations.

This study examines how Arab media have engaged with the United Nations Sustainable Development Goals (SDGs) over the past decade and evaluates the alignment between media coverage and official government priorities. The research addresses the lack of large-scale, Arabic-focused analyses in SDG discourse, which is often dominated by English-language studies. We collected and processed a unique dataset of over 1.2 million Arabic news articles from ten countries between 2010 and 2024. Using a combination of data augmentation, deep learning (specifically, Transformer-based models), and large language models (LLMs), we trained classifiers to detect references to the SDGs and categorize articles by specific SDGs. The results reveal regional patterns in SDG coverage, with North African countries focusing more on governance-related goals, while Gulf countries emphasize economic and environmental themes. Our findings reveal a general alignment between media discourse and official SDG priorities, with notable exceptions. This study is the first to combine artificial intelligence (AI) methods and Arabic media at this scale for SDG analysis, offering new tools and insights for policymakers, media professionals, and development stakeholders.

In recent years, filter pruning has been recognized as an indispensable technique for mitigating the significant computational complexity and parameter burden associated with deep convolutional neural networks (CNNs). To date, existing methods are based on heuristically designed pruning metrics or implementing weight regulations to penalize filter parameters during the training process. Nevertheless, human-crafted pruning criteria tend not to identify the most critical filters, and the introduction of weight constraints can inadvertently interfere with weight training. To rectify these obstacles, this article introduces a novel mask learning method for autonomous filter pruning, negating requirements for weight penalties. Specifically, we attribute a learnable mask to each filter. During forward propagation, the mask is transformed to a binary value of 1 or 0, serving as indicators for the necessity of corresponding filter pruning. In contrast, throughout backward propagation, we use straight-through estimator (STE) to estimate the gradient of masks, accommodating the non-differentiable characteristic of the rounding function. We verify that these learned masks aptly reflect the significance of corresponding filters. Concurrently, throughout the mask learning process, the training of neural network parameters remains uninfluenced, therefore protecting the normal training process of weights. The efficacy of our proposed filter pruning method based on mask learning, termed MLPruner, is substantiated through its application to prevalent CNNs across numerous representative benchmarks.

Plagiarism detection in Arabic texts remains a significant challenge due to the complex morphological structure, rich linguistic diversity, and scarcity of high-quality labeled datasets. This study proposes a robust framework for Arabic plagiarism detection by integrating Siamese neural networks (SNN) with state-of-the-art transformer architectures, specifically AraT5 and Longformer. The system employs a hybrid workflow, combining transformer-based encoders and a classification objective to implicitly learn textual similarity. To address the inherent imbalance in Arabic plagiarism datasets, both weighted cross-entropy loss and Dice loss functions were utilized to optimize model training. Extensive experiments were conducted using the ExAraCorpusPAN2015 dataset, demonstrating the effectiveness of the proposed architecture. Results indicate that AraT5 with weighted cross-entropy loss outperformed other configurations, achieving an F1-score of 0.9058. Additionally, comparative analysis with existing methodologies highlights the superiority of our approach in handling nuanced semantic and structural variations within Arabic texts. This study underscores the importance of transformer-based architectures and class-specific loss functions in enhancing plagiarism detection accuracy in under-resourced languages like Arabic.