AN INTELLIGENT DATA CLUSTERING APPROACH FOR OPTIMIZING ONLINE TEACHING EFFECTIVENESS: A DATA MINING PERSPECTIVE

Abstract

The expansion of online teaching and learning in the digital era has created both opportunities and challenges for maintaining educational quality, particularly in addressing the diversity of learner behaviors. This study aimed: 1) To develop and validate the application of intelligent clustering methods to enhance the effectiveness of online teaching through educational data mining techniques; 2) To identify five distinct patterns of learner behavior using a K-means algorithm enhanced with density-weighted mechanisms and dynamic centroid adjustment; and 3) To design and evaluate the effectiveness of an Intelligent Clustering-Based Teaching Optimization Model (ICTOM) for translating clustering results into actionable, personalized interventions. Data were collected from 1,248 students across three universities during the first semester of the academic year 2066, incorporating indicators of engagement, social interaction, and learning achievement. The results revealed that 1) The developed model demonstrated high quality, with a Silhouette coefficient of 0.72, Gap statistic of 0.89, and Davies-Bouldin Index of 0.83, reflecting the statistical reliability of the clustering; 2) Learners were categorized into five groups: active participants, passive learners, irregular participants, at-risk students, and self-directed learners; and 3) The use of the Random Forest model achieved over 85% accuracy in predicting learning outcomes. Personalized instruction improved overall learning performance by 27.4% and increased learner satisfaction by 34.8%, with the most significant improvement observed among at-risk and irregular participants. This study introduces the ICTOM framework, which systematically links intelligent clustering with personalized teaching strategies.