Analysis Of The Influence In Social Network Evolution

Social Network Analysis is a kind of social research method which is focus on the relationship between the social entities. There are mainly two types of hot topics on social network: influence analysis and network evolution. Influence analysis can be applied in marketing, and research how to use the limited resources to select some influential individuals to test the free samples so that they can finally influence the largest number of people with the strategies of “word-of-mouth” and “viral-marketing”, which is defined as influence maximization problem, when study this problem, we should get the whole network topology, but the network topology is changing in reality, so it is a meaningful task to study the influence of the evolution networks.In this paper, we research the influence maximization problem in static network, introduce some related theory knowledge about it and two kinds of general influence spread models which are linear threshold model and independent cascade model, we also traverse the existing influence algorithm that are greedy algorithm and MaxDegree algorithm and summarize their shortcomings. Then a community-based diffuse efficiency algorithm is proposed to solve influence maximization problem according to the property of community that the connections of the inter-community are denser than that of the intra-community, the experiments results show the community-based diffuse efficiency algorithm performs better than others in the aspect of diffusion degree and time.Finally, we extend the research to the dynamic networks, so the models must be improved to adapt the evolution, and then events based on communities and individuals are proposed, through the analysis of these events, we define two measurements indexes which are social index and influence index, at the same time we introduce a visualization tool of network evolution that is alluvial diagram, which can be understand the evolution of the community over the time intuitively. The experimental results show that the effect of social index is superior to influence index at the initial diffusion stage, however, when diffuse to the bottleneck, nodes of influence index can break the bottleneck faster than nodes of social index so that they can infect more nodes in the network.