Mining the influencing factors of college students’ online learning behavior is of great significance for optimizing online education. Based on social cognitive theory, this paper constructs a model of influencing factors of MOOC learners’ learning behavior and mines the correlation among factors. Multi-attribute event sequence analysis of learning behavior data is carried out using feature extraction technology. The principle of minimum description length (MDL) is introduced in MinDL algorithm to balance the information loss and interpretability of pattern extraction. The K-means-CE clustering algorithm is proposed to achieve the initial center search through crown clustering and determine the optimal number of clusters by combining with the “elbow value”, so as to realize the efficient classification and group portrait of learning behaviors. The MOOC platform learning behavior data is used as the research object to analyze and explore the categorization of students’ behaviors. The results show that the correlation between the seven learning behaviors reaches up to 0.83, and all of them are negatively correlated with the dropout results. The students were clustered into 4 categories according to the characteristics of learning behaviors, which were simple experiencer, good questioner, active explorer, and school bully type.
