Fuzzy Modularity Based Louvain Algorithm For Overlapping Community Detection

Abstract

Due to the fact that things may belong to multiple classifications, the traditional community partition algorithm cannot meet increasingly changing demands; the existing overlapping community partition algorithm still requires specific parameters in advance due to its poor stability. In this paper, a proposition for an overlapping community based on Louvain algorithm is proposed. By introducing fuzzy modularity and calculating the membership matrix of the node for each community, the calculation of the membership matrix is optimized by using the local gradient descent algorithm, so that the node can be assigned to multiple communities simultaneously. With higher accuracy, solid stability and shorter running time (when compared with the original Louvain algorithm), the algorithm in the larger community can be safely applied to the fields of recommendation or prediction.