Research On Overlapping Community Detection In Attributed Networks

Community detection is a fundamental and widely-studied problem.Most existing community detection methods focus on network topologies,but with the proliferation of rich information available for entities in real-world networks,it is indispensable to capture the rich interaction between the structure and attributes in the graph for community detection.By summarizing and analyzing the existing methods of community detection methods,this thesis focuses on two aspects: overlapping community detection based on structure and attribute view and overlapping community detection based on colored random walk,and acquires the following research results:1.We proposed an overlapping community detection algorithm combining structure and attribute view(Combination structure and attribute view for Overlapping Community Detection,COCD).This algorithm consists of three phases,first the attributed graph is embedded into a low-dimensional space by combining the structural similarity,which can reflect the network structure and maintain the network attributes simultaneously,then the structure and attribute view is obtained.Then based on the structure and attribute view,it is possible to acquire the weight of two levels and exact subspaces.Finally,an additional step is added in the community detection process which using weight values to automatically calculate the relative importance of dimensions in different communities.This algorithm can not only automatically balance the topological structure and attribute information of each node,but also capture the attribute subspaces of each community,which improves the quality of community detection.2.We proposed an overlapping community detection based on colored random walk.(Overlapping Community Detection based on Colored random walk,OCDC).This model is able to conquer the limitation of random walk-based community detection methods,which directly utilize the original network topology.Specifically,we firstly select initial seed nodes in the network;Secondly,a seed replacement strategy is developed to capture a better-quality seed replacement path set.Thirdly,the structure-attribute interaction node transition matrix is generated to perform the colored random walk in order to obtain the colored distribution vector;Finally,based on the combination of structure and attribute,the parallel conductance is captured to expand the community.Finally,a large number of experiments are carried out on synthetic datasets and realworld datasets.The experiments show that the two methods proposed in this thesis are better than the existing methods for community detection in attributed networks,verifying the effectiveness and application value of the algorithm in this thesis.