Research On Semantic Community Detection Method Based On Topic Influence Percolation

Community detection in semantic social networks is an important and difficult problem in online social network analysis,which has attracted extensive attention of researchers in various fields in recent years.Most of the existed methods generate communities based on the tendency of users to topics,ignoring the influence of topic itself and the influence and role of topic communication in community detection and the general topology of the generated semantic community is scattered and has a poor interpretability.In order to solve this problem,we will carry out the research on semantic overlapping community detection method based on topic influence seepage.To sum up,the main work of this paper includes:1.The representation of node coordinates in the topic propagation space.Using latent dirichlet allocation as the topic generation model to sample the data.The K topics in the document set contained in the node are used as the basis of k-dimensional semantic space.Calculate the sum mean of the coordinates of all keywords in the document set contained in the node in the k-dimensional semantic space to form a vector,which is used as the coordinate representation of the node semantic information to the semantic space.2.The topic influence seepage intensity function is constructed.The topic influence seepage differential equation is constructed by using the instantaneous point source function in seepage mechanics.The topic influence seepage intensity is quantitatively characterized by solving the equation.3.Nash equilibrium strategy is used to divide communities.A hybrid strategy based on Nash equilibrium is proposed.The propagation process of topic influence is modeled by game theory.The seed node with the largest topic influence is selected as the initial non-equilibrium node of influence seepage,and a semantic community detection algorithm based on seed expansion is proposed.4.Experimental results and analysis.The number of topics in the semantic community,the influence scope of infiltration source nodes,community size and semantic modularity are selected as indicators.The performance difference between this algorithm and classical community discovery algorithm and semantic community detection algorithm on synthetic network and real network is verified.The results show that this method can complete the task of community detection in a short time,has high semantic modularity and semantic cohesion,and is good at discovering community structures with small size and large number.