Yemen Science يمن ساينس: الشبكة اليمنية للعلوم والبيئة، موقع يهتم بأخبار العلوم والتكنولوجيا والصحة والبيئة والسكان
مقالات مشابهة
SANAA, Yemen — Dynamic web usage mining techniques are increasingly replacing traditional data mining approaches in personalized e-learning systems, enabling educational platforms to adapt more effectively to changes in student behavior over time, according to a new systematic review by researchers at Sana’a University.
The review, conducted by Emad Mahmoud Al-Azazi and Ahmed Sultan Al-Hegami from the Faculty of Computer and Information Technology, analyzed research published between 2010 and 2025 to examine how web usage mining has evolved to support personalized learning.
The researchers reviewed 187 peer-reviewed studies selected from an initial pool of 847 publications, focusing on techniques that analyze learners’ interactions with online educational platforms to generate personalized recommendations and adaptive learning pathways.
Unlike conventional systems that rely on historical user behavior, dynamic mining methods continuously update behavioral patterns as new data become available, allowing platforms to respond to shifts in learners’ interests, engagement and learning progress.
The review found that techniques such as incremental mining, sliding window analysis and temporal pattern mining have gained prominence because they can process continuously evolving data without repeatedly analyzing entire datasets. These approaches are better suited to modern e-learning environments, where learner behavior changes throughout a course.
To organize the field, the authors proposed a five-layer framework covering data collection, preprocessing, pattern discovery, dynamic mining, and recommendation systems. They said the model provides a structured view of how educational platforms transform user interaction data into personalized content and learning recommendations.
The review also highlights growing interest in hybrid recommendation systems that combine collaborative filtering, content-based methods and artificial intelligence models to improve recommendation quality while adapting to changing learner behavior. Deep learning techniques, including Long Short-Term Memory (LSTM) networks, transformer models and graph neural networks, are increasingly being explored despite higher computational demands and challenges in explaining their recommendations.
The researchers identified several challenges that continue to limit large-scale deployment of dynamic web mining in education. These include scalability, concept drift as learner behavior evolves, data sparsity, privacy concerns and the lack of standardized measures for assessing the educational value of discovered behavioral patterns.
The review also argues that many studies emphasize technical performance metrics such as prediction accuracy while giving less attention to educational outcomes, including learning gains, student engagement and course completion. The authors call for future research to evaluate whether improvements in algorithmic performance translate into measurable improvements in learning.
According to the study, the field has shifted markedly since 2023 toward real-time and adaptive mining approaches as educational platforms generate increasingly large volumes of learner interaction data. The authors conclude that future personalized learning systems are likely to integrate dynamic mining techniques with explainable artificial intelligence to improve both adaptability and transparency. The study was published in the Sana’a University Journal of Applied Sciences and Technology in June 2026.
