This study evaluates the effectiveness of a Drosophila melanogaster-based virtual laboratory in enhancing research skills and student engagement in genetics learning. The objective was to provide an accessible, technology-driven learning environment that supports the development of critical scientific competencies. Materials included a web-based virtual laboratory designed under a Project-Based Learning (PBL) framework, simulating key genetics activities such as strain identification, culture preparation, crossbreeding, and data analysis. The study used a quasi-experimental design with 122 biology education students, divided into a control group (PBL only) and an experimental group (PBL combined with the virtual laboratory). The researchers collected data using a research skills test, an engagement questionnaire, and observation sheets. Results showed that students in the experimental group significantly improved in formulating research questions, designing experiments, analyzing data, and drawing evidence-based conclusions compared to the control group. They also reported higher motivation and satisfaction, and observations confirmed greater active participation, accurate procedure execution, and effective tool use. These findings demonstrate that virtual laboratories can meaningfully support contextual and sustainable scientific skill development. The implications for future research include expanding virtual laboratory designs across broader scientific domains, integrating adaptive learning features to personalize student experiences, and conducting longitudinal studies to examine long-term impacts on scientific reasoning and professional preparedness. By emphasizing digital technologies within participatory pedagogy, this study highlights the potential of virtual environments to transform genetics education in digitally mediated contexts.
Intelligent traffic monitoring is a key element in smart transportation systems, and the ability to perform a fast and secure analysis of heterogeneous visual information is crucial for timely traffic control and incident handling. While complementary drone- and satellite-based perspectives provide local and global understanding, current approaches to fusion based on Edge AI typically rely on static fusion strategies, have limited adaptability to variable traffic conditions, and are vulnerable to attacks that compromise inference integrity. In response to this, the paper introduces an Adaptive Context-Aware Secure Edge Reasoning Framework (ACSERF) to overcome these challenges in intelligent traffic monitoring through drone and satellite image analytics. Both the proposed cross-scale reasoning mechanism and cyber-confidence are dynamic models. The former models cross-scale spatial relationships between drone and satellite observations dynamically, while the latter estimates the cyber-confidence for the reliability of visual information before inference at the edge. Moreover, an edge optimization strategy that adaptively allocates resources allows computational loads to be adjusted, enabling offloaded computation processes to run at low latency without compromising their detection capability. It was tested with available/available drone and satellite traffic image datasets and tested under a variety of traffic densities and environmental conditions. From experimental results, an overall accuracy of 99.03%, precision of 98.85%, recall of 98.71%, F1-score of 98.78%, and mAP of 98.90% were achieved while reducing inference latency by 30.9% and computational overhead by 26.4% over the previously reported methods. The cybersecurity module also achieved a 98.12% detection rate and a 1.57% false alarm rate, enhancing its ability to withstand data-injection and adversarial attacks. These results show that the proposed ACSERF can be an efficient, scalable, and cyber-resilient solution for next-generation intelligent traffic monitoring in smart transportation environments.
Marital satisfaction is one of the major determinants of personal well-being and family stability, but the predictors are still complicated in relation to cultural contexts. This paper aims to identify and model the variables that influence marital satisfaction using machine-learning-based predictive modeling and Maslow's needs. The preprocessing of a global survey dataset involved variable mapping, duplicate removal, type correction, and range reversal to ensure consistency. Exploratory and correlational analyses showed that marital satisfaction was strongly correlated with psychosocial variables, including love, safety, and self-actualization. Feature selection (Recursive Feature Elimination) found six important predictors: age, length of marriage, number of children raised, religiosity, safety, and love. It was trained on five machine learning models, the best test accuracy of which was 86.84% with the ensemble method Gradient Boosting using a binary classification scheme, with repeated hold-out validation and statistical testing used to assess evaluation robustness. Although there were challenges due to class imbalance, binary grouping achieved better predictive performance than multiclass classification. These results indicate that methods integrating psychological frameworks and machine learning may yield valid, interpretable models of marital satisfaction, with potential applications in relationship counseling, cross-cultural studies, and educational programs.
Classical Arabic narratives present unique challenges in the computational analysis of their texts, due to the language's highly complex morphology, orthographic ambiguity, and stylistic variations over time. This study presents a framework for Explainable Artificial Intelligence (XAI) for cognitive and pragmatic study of Arabic texts with two typical tasks, namely author attribution and Named Entity Recognition (NER). The ground truth used in the framework is not synthetically created, but rather comes from two already available resources, that are sufficiently rich to provide strong, machine-verifiable ground truth: the OpenITI corpus segmented in passages and labeled by the Author URI for multi-class attribution for six classical authors and CLEAN-ANERCorp, a corrected CoNLL format Arabic NER corpus recast as a windowed-context classification task with labels PER, LOC, ORG, MISC and O. The four traditional machine-learning classifiers, Logistic Regression, Multinomial Naive Bayes, Linear Support Vector Machine (SVM), and Random Forest are systematically benchmarked after the normalization and TF-IDF vectorization of the Arabic texts (unigrams and bigrams, vocabulary limited to approximately 8,000 features, minimum document frequency of 1). Models are tested using an 80/20 stratified train-test split, and accuracy, weighted precision, weighted recall, weighted F1-score, Matthews Correlation Coefficient (MCC), and Cohen's Kappa are used to evaluate models. The Linear SVM achieves the best performance with an 80/20 split, with an author-attribution F1 of 0.932 and an NER F1 of 0.809, while an 80/40 split yields an author-attribution F1 of 0.942. The ten most important TF-IDF features are extracted for each class and presented as a pair of Arabic-English word lists, thereby providing algorithmic transparency. Together, the results demonstrate that linear discriminative models and appropriate lexical feature engineering can provide a very accurate and interpretable baseline for the development of natural language processing algorithms for Arabic.
Breast Cancer (BC) remains one of the leading causes of mortality among women worldwide. Early detection plays a crucial role in enabling timely diagnosis and effective treatment, thereby significantly improving patient survival rates. However, conventional breast cancer detection techniques are often time-consuming, less accurate, and heavily dependent on human expertise, which increases both operational costs and the likelihood of diagnostic errors. To overcome these limitations, Machine Learning (ML)–based approaches offer a promising alternative by reducing human intervention, lowering costs, accelerating diagnosis, and improving predictive accuracy. In this work, nine ML algorithms are employed for the automatic detection and classification of breast cancer, namely Random Forest (RF), AdaBoost, Gradient Boosting (GB), XGBoost, Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Decision Tree (DT), and Neural Network (NN). The experiments are conducted using the publicly available Breast Cancer Wisconsin Dataset (BCWD) obtained from Kaggle. The proposed framework systematically evaluates and compares the performance of these algorithms, while also highlighting their respective strengths and limitations in the context of breast cancer diagnosis. The experimental results demonstrate the effectiveness of ML models in breast cancer detection and classification. Among the evaluated algorithms, Logistic Regression achieved superior performance across multiple evaluation metrics. Specifically, LR achieved an accuracy of 99.12%, precision of 99%, recall of 99%, F1-score of 99%, AUC of 99%, a loss of 0.32%, and an ROC of 1. Performance assessment is conducted using accuracy, precision, recall, F1 Score, loss, AUC, and ROC curves across all models. Furthermore, a comparative analysis of the nine ML algorithms is conducted to validate the robustness of the proposed approach. To assess model generalization and practical applicability, a Streamlit-based web application is also developed, enabling real-time evaluation and reliability testing of the trained models.
