Research On Network Influence Propagation Algorithm Based On Community Structure

Recently,with the popularity of Internet technology, people's accesses to gather information is changing form traditional broadcast, television and newspaper to microblog,post bar and moments. Online social network is continuously combining with traditional social network, which generated massive data. This offers unprecedented opportunities for the analysis of social network. Massive researchers have deeply studied multiple problems in this field, such as influence maximization in social network and the law of information propagation. The problem of selecting the top-k influencing nodes and constructing accurate influence propagation model become hot research topics in the field of social network. Based on the deeply analysis of previous works, this paper first proposes an enhanced centricity algorithm by importing The Strength of Weak Ties theory aiming at the problems of current influence maximization algorithms in social networks. Second,according to the detailed analysis of linear threshold model,this paper proposes a new propagation model in social network. The concept of node connection strength and information's own attraction properties are introduced by merging the differences of nodes in social networks. The detailed research contents are as follows:(1) Community Structure Based On Key Node Centricity Algorithm. In order to combine the properties of network structure, this algorithm evaluates nodes' influence by regarding both boundary nodes and nodes in community as key nodes. According to The Strength of Weak Ties theory,networks with strong cohesion are not benefit to nodes to gather outer information. So examining the attributes of boundary nodes which are the weak inter-community connection is benefit to the inter-region information propagation. What's more, selecting the nodes with maximum influence in the community may cause information to propagate rapidly within the community. Combing them is benefit to the information propagation in the whole network. This paper verifies algorithm's effectiveness in three different aspects on three datasets.(2) Connection Strength Threshold Model. According to the deep research on linear threshold model,this model assumes for a particular node,the influence from its neighbor nodes at the same time is equal. However,in real social network,different individuals may have different closeness's. The differences among individuals result in the differences for node being influenced by its neighbors. What's more, the effect of information propagation is closely related to information's own attractive. Therefore,this paper proposes a LT model based connection strength threshold model. This model introduced the concept of node connection strength and information's own attraction properties by inducing the advantage of linear threshold model and merging the differences of nodes in social network, which enhanced the perimeter buv in linear threshold model and propose new formula for node's influence.