Identifying High-Level Student Behavior Using Sequence-based Motif Discovery

We describe a data mining technique for the discovery of student behavior patterns while using a tutoring system. Student actions are logged during tutor sessions. The actions are categorized, binned and symbolized. The resulting symbols are arranged sequentially, and examined by a motif discovery algorithm to detect repetitive patterns, or motifs, that describe frequent tutor events. These motifs are examined and categorized as student behaviors. The categorized motifs can be used in real-time detection of student behaviors in the tutor system. Computer Science Department, University of Massachusetts, Amherst