Influence Analysis Of Online Social Networks

With the development of science and technology and frequent interaction in a network, social networking platform is becoming more and more popular, so it’s a powerful means of human communication and information transmission. In a social network, the interaction behavior of the users will produce huge amounts of data,These data not only provide convenience for the study of influence in social network,but also make the research of influence becomes more complex. In the era of big data,the complexity of the social network is widely discussed in academia. Therefore, this paper will conduct detailed studies from two aspects of the diffusion rule and mining opinion leaders.First of all, through the analysis of the communication process based on SIR, we put forward a kind of reverse diffusion model based on immune point and describe reverse effect in the process of diffusion. Diffusion regularity of the new model can effectively solve the effect of the immune point at the differences between reality and model. Finally, we will have a simulation experiment in Netlogo and analyze and compare the results of the new algorithm and the traditional algorithm based on SIR model. The experimental results show that the new algorithm can present different sizes of slow rate under different parameters so that hinder the entire spread of influence.Secondly, on the basis of the network topology, we put forward the concept of the secondary average connectivity, which used to measure the complexity of neighbor nodes in the network. At the same time, we use the measurements as a standard of mining opinion leaders and write opinion leader mining algorithm. Finally,the algorithm will be implement by MATLAB and we will compare results between three kinds of measurement of degree centrality, betweenness centrality and Pagerank with NodeXL analysis tool. This experiment shows that the evaluation standard based on the secondary average connectivity can effectively mining the opinion leaders in the network. So the experiment confirm the effectiveness and superiority of the proposed algorithm.