A Potential Overlapping Community Detection Algorithm for Learners Based on High-order Organization

Learner network topology is sparse and traditional community detection algorithms cannot detect their potential communities for inert/cold start learners. To address this type of problems, a potential overlapping community detection algorithm (POCDL) for learners based on higher-order organizations is proposed. The POCDL algorithm is a local graph clustering algorithm that firstly uses the friend, classmate and teacher-student relationships in socialized online course platforms to solve the sparse learner network data problem, and then mines the higher-order organizations in the learner network and reconstructs the learner network; finally, the initial seed set is selected based on the degree centrality of learners, and local community detection based on community belongingness and community closeness. Experimental results on artificial networks and SCHOLAT dataset show that the POCDL algorithm can detect communities better for inert/cold-start learners, and also has some generalizability to other types of complex networks.