Digital Intelligence-Driven Analysis of English Learning Behaviors and Personalized Teaching Model Construction

Abstract

Digital-intelligence technologies are reshaping language education by enabling large-scale behavioral sensing, adaptive modeling, and individualized instructional feedback through networked learning environments. From an engineering perspective, the reliable transmission and analysis of learner-interaction data share common concerns with intelligent communication and signal-processing systems, including latency, data integrity, and adaptive classification. T his s tudy c onstructs a four-module personalized English teaching model covering learner profiling, behavioral analysis, instructional decision-making, and adaptive feedback regulation. Based on platform logs, questionnaires, and academic -performance records from 480 undergraduate learners, K-means clustering, association rule mining, and randomforest modeling were used to identify learning-behavior patterns and predict instructional needs. Four learner clusters were identified, and exercise completion rate and interactive engagement emerged as the strongest predictors of learning outcomes, with five behavioral features explaining approximately 79% of the variance in final examination scores. The classification model achieved 87.3% accuracy and an AUC of 0.924 under cross-validation. A 16-week controlled experiment further showed that the experimental group achieved a 21.1% score gain, compared with 8.8% in the control group, confirming the practical effectiveness of the digitally intelligent personalized model.