Identifying At-Risk Learners in Engineering Education: A Self-Regulated Learning Perspective Based on Digital Traces

In the context of digital transformation in engineering education, precisely identifying at-risk learners through learning analytics has become essential for enhancing teaching quality. Based on Self-Regulated Learning (SRL) theory, this study extracts a six-dimensional feature vector from students' online behavioral traces. Using the K-Means algorithm, four distinct strategic prototypes were identified. This paper specifically focuses on the "Effortful Low-Achiever" and "Surface Participant" profiles to develop an early-warning model based on centroid distance. Results indicate that the model effectively identifies students with potential academic risks during the midterm stage, providing an empirical foundation for precision intervention.