The Study Of Influence Maximization Algorithm Based On Community Structure

In recent years, due to the excellent performance in expanding social circle, lookinglost contact with friends and to maintain contact with friends among other aspects, somelarge social networking sites such as FaceBook, Twitter, etc. got a great success. socialnetwork has become a huge proliferation of news and product marketing platform,Because it can make information technology in the network quickly affect people. In orderto tap the potential of social networks as a platform for information diffusion, there aremany problems to solve. Influence Maximization is an important issue. Based on thefeature that the information is easy to propagate in nodes of the same community in realsocial networks, a new influence maximization algorithm based on community structure isproposed. This paper studies as following:Firstly, a new algorithm called IDP for community detection is proposed in this paper.In the label propagation process of partition, it is mainly made a reasonable set in threeaspects, node’s Neighbor Set, nodes’ traversal sequence and label’s Update Rule, based onconsidering the inherent characteristics of network, Compared with CGA-IM, the proposalof combination entropy in combination will combine the communities.Secondly, To evaluate the performance of IDP, we propose two new metrics, NIDIand SPCI, according to the feature of community partition in based information diffusion.We conduct IDP and the baseline algorithm CGA-IM in different scale datasets andevaluate the performance of these two algorithms with NIDI, SPCI, SI and RTI.Thirdly, Based on the above results, a algorithm of influence maximization based oncommunity detection is proposed, the main idea is to scale sparse network re-calculation,Compared with greedy algorithm in influence range and running time.Finally, the experimental results show that the efficiency of the algorithms of IDP andCIP.