Community detection by fusion of motif-aware and graph Transformer encoding

GUO Xing-jun, LI Xiao Hong, SHIWan-yao, GAO Wen-chao

Abstract: The higher-order connectivity structure has been largely ignored, which contains a bettersignature of community compared with the lower-order connectivity structure,and the high-order infor.mation causes the inevitable fragmentation problem, To solve those problems,a motif-aware and graphTransformer(MGTrans) for com munity detection is proposed. Firstly, the maximal complete subgraphin the graph is searched and regarded as a motif, and the original graph is reconstructed with the motif asa unit to capture the motif adjacency matrix, At the same time, mixed-order outer-cut edges encoding isused to obtain the residual edge information of the original graph to solve the fragmentation problem ,and position information and edge information on the reconstructed graph are captured through a position encoding matrix and motif short path with weight encoded. Then,the initial features are extractedby a graph transformer. Combing position encoding matrix, edge encoding matrix and initial featuresthrough the attention network to get motilf embedding matrix for the community detection. Finally, Theexperimental results on several diferent datasets show the effectiveness of the MGTrans in improvingthe community detection performance of stateof-theart methods and effectiveness for overlapping com-munity detection and multi-community public node detection.