This study focuses on the innovative application of artificial intelligence technology and teaching transformation in English education in colleges and universities, constructs a three-level theoretical framework (student model, teacher model, domain model) for intelligent computer-assisted instruction ICAI system, and explores the data-driven teaching optimization path by combining the time-series clustering and decision tree algorithms. Based on the real data from the intelligent teaching platform of a university, the study clusters students’ learning behaviors through the time-series k-means algorithm KmL, identifies four types of differentiated learning groups, efficient learners (N=625), task-oriented (N=3011), passive participants (N=4276), and passive groups (N=247), and reveals their behavioral characteristics in resource use, interactive participation, and other 10 behavioral characteristics in the dimensions of resource use, interaction participation, etc. The decision tree algorithm was further utilized to mine the academic performance association rules, and found that classroom mastery, listening time and vocabulary were the core factors affecting the performance, such as the Rule 1 confidence level of 61.23%. The study shows that the data-driven ICAI system can realize the dynamic adaptation of teaching strategies, provide technical support for personalized teaching and precise intervention, and promote the transformation of English education in colleges and universities to intelligence and refinement.
