Research On Community Detection And Evolution Analysis In Social Interaction Network

With the increasing development of information technology,network science has been paid much attention.Network science is the integrated disciplines,which has a variety of applications in computer networks,social networks,and biological networks and so on.Because of the rapid development of communication devices,social interaction network obtains the much largest progress.Since the social interaction network contains rich social information,which is significant to analyze the life custom of people.Thus,the research of social interaction network needs more attention.Community structure is ubiquitous in the social interaction network.The members of the same community have the common properties in some aspects,which is significant to analyze the social activities.We will introduce the paper's content as follows:The community structure is usually overlapping and non-isolated.Firstly,we research the overlapping community detection.The line graph of original network is constructed by making use of the line-graph theory.Then,we exploit topic-finding-model to detect communities.In this paper,we show that the overlapping community partition is fuzzy;there is not a specific definition for overlapping community.We propose a method that the overlapping degree can be adjusted,and obtain the result of community partition at last.Social interaction network between people own a lot of diversity,which is called the multi-view social interaction network in this paper.Secondly,we research the community of multi-view social interaction networks.Unlike the single-view network,the conventional mathematical model is difficult to describe the multi-view network completely.Furthermore,it is not necessary to take the weight between multi views into consideration.In this paper,we use the tensor to describe the multi-view networks,and formulate the optimal model of tensor decomposition.Compared with HOOI algorithm and Cr NC algorithm,we find the Cr Nc algorithm has better performance.Thirdly,we research community evolution in dynamic networks.Unlike the static network,dynamic network structure changes over time.Community evolution,generally speaking,is the communities throughout the changes in all timeline.However,the existed schemes do not take the noise into consideration.In order to solve this problem,we analyze community detection and community evolution separately.Finally,we use the proposed algorithm to obtain the evolutionary community structure.In this paper,we research the community detection and evolution analysis in social interaction networks,and obtain community partition in different conditions.The numerical results demonstrate the proposed schemes can achieve better performance than the existed schemes.