| Social online learning is widely considered to be a new educational revolution and is a popular topic of research in various subject areas.In recent years,with the increasing amount of available data in large-scale complex networks,research related to community detection of complex learner networks has gradually become a popular topic,but there are two main challenges in community detection of learner networks: first,the topology of learner networks is sparse compared to general complex networks,and traditional community detection algorithms cannot identify potential communities for inert/cold-start learners The second is that the variety of entity relationships in learner networks is complex,and learners may be in different learning circles（communities）at the same time,making it difficult for existing algorithms to simultaneously balance detection quality and time complexity for learner network community detection.This paper focuses on learner networks,aiming to identify overlapping potential communities in learner networks and discover the cohesive structural features of groups in learner networks,so as to reveal the organizational behavior of complex systems described by learner networks,with the main research contents including.（1）In-depth analysis of the topological characteristics of learner networks and the influence of higher-order organizations on communities,a potential overlapping community detection algorithm POCDL for learners based on higher-order organizations is proposed.aiming to improve the quality of learner community detection,the core idea includes three stages: reconstructing learner networks using triangular moduli,local community detection,and community optimization.First,the learner network is reconstructed using four modal organizations to form a learner edge-weighted network;second,the influence of nodes is calculated using degree centrality and arranged in descending order,and the nodes with the highest degree centrality are combined with the modal to form the initial community,followed by local expansion to form the initial community detection results;finally,a community closeness evaluation model is constructed for community optimization,where neighboring communities are judged to be merged using the community closeness Finally,the community detection results are obtained by constructing a community closeness evaluation model for community optimization,where neighboring communities use community closeness to determine whether to merge,and isolated nodes use the proximity principle to select communities.The effectiveness of the algorithm is verified by comparing it with that of a single modal on the scholars’ network dataset;at the same time,comparison experiments are conducted with four classical overlapping community detection algorithms on two groups of artificial networks,and the results show that the algorithm POCDL can effectively identify community detection while solving the problem of potential learning community detection for learners.（2）With the help of the approximately linear time complexity of label propagation,a label propagation-based overlapping community detection algorithm for learners,OLCDA,is proposed.aiming to improve the efficiency and quality of community detection in large-scale learner networks,the main steps of the OLCDA method are,firstly,to design a learner node importance model to calculate the node importance of the learner network and obtain the propagation order of node labels accordingly.Then,the maximum node favorability metric is used to calculate the connections among learners.Finally,the combined influence of learners’ neighbors is calculated.The iteration is stopped until the dominant label is unchanged.The results of extensive comparison experiments on 10 real network datasets and 4 sets of artificial network datasets show that the OLCDA algorithm is not only able to identify learner communities faster also has high quality of community detection,in addition,it has generalizability to community detection of other types of complex networks. |
