Research On Community Detection Algorithm And Community Profile Based On User Call Record

Recently,with the increasing popularity of mobile phone and smart terminal in China,there has generated a large number of data examples of mobile user.Call record data can reflect social relation of mobile user in the real world.In the network the mobile user's social circle is called the community.Analyzing call records can find that mobile users may be in different communities,such as relatives and friends' community,work community and hobby community.Different communities have different characteristics,these characteristics are called community profile.In this paper,we focus on two questions based on the call record data.One is how to find the community in the call record,and the other is how to construct the community profile for the discovered community.More specifically,for the first question,we first propose a weighted edges based modularity evaluation criteria,and then a local community detection algorithm by using our proposed evaluation criteria.For the other question,we propose a method of constructing local community profile of call record network.In addition,in order to accurately learn the preferences of each mobile user on multiple perspectives,we also propose a framework to construct local community profiles from multiple perspectives.Experimental evaluation on both the open data sets and call record network data set verifies the effectiveness of our proposed local community detection algorithm in the following two cases.(1)For labelled network,the algorithm can find a completely real community.(2)For the unlabelled network,the local community discovered by our algorithm has a high degree of modularity.Moreover,the experimental results on the call record network indicate that our framework can effectively learn the call habits andpersonal preferences of mobile users.