Research Of Community Detection Algorithm Based On Trust Model In Social Network

With the repaid development of the Internet technologies,the social network has become the main platform for people daily communication,personal life display and message release.Community detection is a hot issue of social network research.Exploring the community structure of the social network will allow a deeper insight into the topology features of the network,also this work can provide strong support for the subsequent study and application in public opinion monitoring,opinion leaders discovering,personalization recommendation and so on.Many scholars have used the graph theory to describe and depict the community structure,and put forward the respective detection algorithm from different angles.However,the problems still exist in the current detection algorithm need to solve,such as excessive reliance on edge,the effect of overlapping community detection is not ideal and so on.At the same time,in the research of social network,more and more scholars have realized that between the users in social network not only exist the dominant relation,but also exist the tacit relation that reflected by friend similarity,attribute similarity and interest similarity and so on.In order to analyze the relation feature of social network more reasonably,many scholars have started using the trust to judge the relation feature by the individual weight,relation strength and so on,also define the trust calculation method and transfer rules to complete the relation description of social network,finally improve the accuracy the social network analysis.Therefore,in this paper,firstly defines a trust calcution of nodes in social network,through using the trust to describe the feature of relation between the nodes in social network,and on this basis comes up with a community detection algorithm,finaly verify the actual performance of the algorithm by comparative analysis.The specific content is presented as follows:1)Put forward a trust calculation method that combining the relation strength and similarity.On the basis of existing research,analyzing and selecting the factors that influence the trust between nodes,defined the trust calculation method of relation trust and similarity trust that respectively produced by relation strength and similarity between nodes in social network.The trust calculation method is the basis of subsequent research.2)Design a Trust-Based Local Overlapping Community Detection Algorithm TLCDA.TLCDA abstracts the social network into data field and uses the trust potential to describe the influence of between in local scope,then completes overlapping community detection by using rough K-Mediods clustering.3)Enact the experimental scheme and complete comparative analysis.We select three different types of network that covers LFR artificial benchmark,classic real network and weibo network.Next,we give the related evaluation index of community detection and complete the effect verification of TLCDA by comparing with the classical community detection algorithm.