Research On The Evaluation Mechanism Of Node Influence In Online Social Network

In recent years, with the broad applications of Web2.0 technology, online social network has developed rapidly, which provide a platform for people to present themselves and share resource. Online social network has large users' population base, which attract a great many scholars to study. This paper focuses on the structure of network and user characteristics, and puts forward two kinds of methods to evaluate the influence of nodes in online social network. The major work of this paper presents as follows:Based on the comprehensive analysis of structural characteristics and user characteristics, the method to evaluate influence is proposed. Firstly, this thesis analyzes the users'attributes and behavior, devotes to users' attribute with the quantitative approaches. The paper uses the analytic hierarchy process to determine the weight of each attribute. Then the idea of PageRank algorithm is introduced which is used to measure the interactions among users' behavior. And then the paper modifies the probability according to online social network properties. Finally, the IR (Influence Rank) evaluation model is proposed which based on users'attributes and behavior. The experimental results show that the IR model can evaluate users' influence with higher efficiency and more rationality than traditional models.The node influence evaluation method based on topological potential is proposed, and the topological field theory is introduced to the online social network. The thesis fully considerate the users' attributes and network topology structure. Be similiar with physical field, there is a virtual field around each node of online social network. Every node in the field will be affected by other nodes. Then the topological potential is introduced which is used to measure the interactions among users. The topological potential function is constructed, which indicates the users' influence. According to the basic characteristics of online social network and topological field theory, we should compute three parameters of topological potential function:node mass, network distance, influence factor.According to result of evaluation, this thesis proposes an algorithm based on node influence for detecting local community structure. In essence, on the basis of comparative result for neighboring nodes' influence, the communities were detected. The large influence nodes will attract the little influence nodes. Firstly, the paper selectively detects community in Breadth First Search in the terms of the largest influence node. Then we can get community structure of entire network by repeating this progress. Finally, the experiment results show that the algorithm is efficient and accurate.