Social Network Community Impact Assessment And Prediction Research

Social network has become a significant platform for people to achieve information dissertation and sharing. With the population keeping expanding, social network has generated enormous impacts and further infiltrated them to all walks of human life. Therefore, the efforts of influence analysis towards all kinds of entities in social network have become the hotspots in current years.Traditionally, the influence analysis is divided into two types in accordance with different subjects of study, namely the different entities of user and community. From the views of influence range and influence degree, community influence would, no doubt, impact more than single individuals in social network since the community is an integral whole which is composed by various users. Therefore, community influence analysis has broader prospect and more positive sense. At present, many researches mainly focus on the influence mining for single individuals,while little critical attention has been paid to community influence evaluation in social network.Meanwhile, most individual influence analysis methods are given based on users' profiles analysis or based on probability and statistics method. The critical facts, including eliminating interference factors(e.g. “zombies”) and the trustworthy relationships among users, have been taken little account in influence analysis. In addition, other important factors, such as community scale and closeness degree of users inside the community, play little roles in influence analysis for community.For overcoming the disadvantages of traditional influence analysis, two models, named community influence evaluation model(CIEM) and community influence predicting model(CIPM),are proposed to evaluate and predict the community influence as a whole. Meanwhile, some related calculation methods and theories are also given. The main works of this dissertation includes:1) In the community influence evaluation model, first, we presented CIEM model framework and related definitions; secondly, we give the method to get rid of zombie fans and incorporate trust into the user influence calculation, and give the user final influence calculation method; furthermore,we integrated the comprehensive influence of community user ? community scale and the relationship of users in community together, and give the calculation method of community influence; in the end, we put forward the calculation method of community extend influence,namely "the community to users in the community", "the commuinty to the users outside the commuinty," and "community to community";2)In the community influence prediction model, first of all we give the framework of CIPM model and its related definition; secondly, we give the calculation method of link prediction reliability; finally, we give the calculation method for community influence prediction;3)Simulation results show that the CIEM model calculation accuracy compared with the weighted average method with high accuracy, CIPM model in the prediction of network security has higher application value; at the same time, the results show that CIEM model and CIPM model alignment is higher, thus further we prove the correctness of the proposed method.