Research On Overlapping Community Detection Algorithm Based On Multiple Network Transform

In the traditional community network discovery,people's goal is to find the non-overlapping community in the network.All vertices can only exist in a single community,and the nodes between communities will not overlap.Moreover,the vertices within communities are closely related to each other,and the vertices between communities are evacuated from each other.However,with the further research,scholars found that the non-overlapping structure community found that the algorithm can no longer meet the needs of people,and the vertex can exist in multiple communities at the same time,namely the overlapping community structure.This kind of community structure is more consistent with the community network existing in the real world,and has more research significance.Traditional overlapping community discovery algorithms mostly take nodes as the main research object,but node-based community division is difficult to deal with the real network with more overlapping nodes,so scholars turn their attention to the edges,and the overlapping community algorithm based on edge division has been greatly developed.Based on the single-layer network mapping(SLG),this paper proposes the overlapping community discovery algorithm of multi-layer network mapping(MLG).SLG transforms the original graph into edge graph,divides the community on the edge graph,and then converts the result into the division result of the original graph.The essence of SLG is to divide the community with the information contained in the edge of the network.However,SLG only goes through the transformation of edge graph once,which will lead to insufficient overlap of results.Only overlapping nodes can be obtained,but no overlapping relationship can be obtained.Moreover,there are few overlapping nodes,which are different from the real network.MLG algorithm mainly has two aspects of improvement: the introduction of multi-layer network mapping idea,in the edge map,N iterations,finally converted to the original map community division results of the overlap degree and overlap range increase;In the process of network mapping,threshold value is set to filter edges with small similarity,so that the edges with small similarity are not mapped to the next edge graph,so as to reduce interference for the subsequent community discovery process.In the execution of MLG algorithm,the network graph is firstly preprocessed to remove redundant edges,and then the original graph is mapped for many times to get the multi-layer edge graph.The multi-layer edge graph is divided by hierarchical clustering algorithm,and the multi-layer edge graph is finally converted into the corresponding division of the original network graph.This article will MLG algorithm comparing with SLG algorithm and CPM algorithm,the experiment using the eight data sets(4 kinds of standard network data collection of Dolphins,help,Politics says,"and four kinds of artificial network data set 128-10-4,40-128-4,128-60-4,256-60-4),and three kinds of classical community assessment method(EQ,PD and CR)detailed analysis with the result of the experiment,the experimental results show that the MLG CPM algorithm and the algorithm is superior to the SLG algorithm,improve the rationality and the veracity of the community.