Research On Community Detection Algorithm Based On The Intimacy Between Users And Density Peaks

The era of social information has started with the rapid development of information technology and the popularity of intelligent hardware.The online social network has changed people`s daily life and entertainment,and a variety of social networking tools appear,such as Microblog,We Chat,Zhi Hu etc.which make people closer,communicate more convenient and fast,and prompt the rapid development of online social networks.Online social network records a large number of information,and the relationships are not exactly the same,so the community tendency is more and more obvious.In order to understand community characteristics and the law of community evolution better,a large number of scholars start to research the social network.Though the method of community detection,social network can be divided into some small community,which will help us understand the network structure more clearly.This paper is focus on the following works about the social network.First,an improved method of measuring user similarity is given.Most of the algorithms found by the community can be used for effective community identification,but the disadvantage is that it only considers the direct and non-directional relationships between nodes,which is unreasonable in a real online social network,as the similarity between nodes cannot be measured precisely only by direct and non-directional relationships between nodes.After the consideration to the direct and indirect relations between the nodes and the influence of the directivity of the relations on the similarity between the nodes,a new method of calculating intrusion based on user relationship comes up.First,we give the generation algorithm of follow and fan matrix,and the definition of direct intimacy and indirect intimacy.Considering the directional relationship of follow and fans,we provide the formula of direct intimacy.Considering the indirect relationship between nodes,we provide the indirect intimacy calculation method.At last,we give the method of calculating the user's intimacy with the calculating process which can integrated measure the structure characteristics between the nodes.Then,the clustering algorithm based on density peak and fast search is improved.As an efficient and novel clustering method,it can automatically identify the size of the community and obtain the cluster structure of arbitrary shape.However,when calculating the distance attribute of a node,it may lead to splitting the same cluster structure into two clusters,which affects the result of the algorithm.In this paper,the clustering idea is applied to the study of community detection in social network.Combining with the characteristics of social network,this paper gives an improved method to identify the community center,so that it can identify the community center more accurately,and we provide the community detection algorithm based on density peak.Then,we combine the two methods to get the user intimacy matrix by the intimacy calculation method based on user relationship,and the importance and distance attribute of users by community detection algorithm which is based on density peak,and make the attribute calculation is more reasonable.At last we provide a complete community detection algorithm based on user intimacy and density peaks.Finally,the results of the algorithm are verified on the microblog dataset and the public dataset.The experiment shows the feasibility of the algorithm and the effectiveness of the algorithm.The algorithm based on parameter adjustment strategy has the advantages of good flexibility and applicable.In the non-directional user relation network,it is proved that the algorithm has better generalization.