Maximizing The Impact Of The Problem-oriented Theme

In recent years, micro-blog, online forums and other social networking sites are becoming more and more popular. So mining the most influential users in the network is greatly important. Viral marketing is a very effective marketing strategy. Influence spread in the social network through personal social relations. People influence their friends, family or co-workers by word of mouth. In this background, the influence maximization problem has become a research hotspot of social networks. There are lots of research results about influence maximization in social network. Influence maximization problem is to select a finite number of nodes forming a seed set to influence other nodes. Ultimately, the final impact maximizes the largest set of nodes on the whole network.A lot of research considered for this problem, such as mountain climbing greedy algorithm. However, previous works have not considered the importance of topic in solving the influence maximization problem in a social network. The fact that the preferences of different users are not the same largely affects the accuracy of the final result.Based on the ordinary influence maximization problem, this article brings forward a new algorithm called GAT (Greedy Algorithm Based on Topic) algorithm to obtain the most influential users under a particular topic. The GAT algorithm first uses latent semantic analysis to calculate the preference value of every users for a particular topic and builds an expanding independent cascade model; on the basis of the new propagation model, GAT adopts hill climbing greedy algorithm to get the seed set which could influences the maximum of nodes. Compared with the previous work, GAT considers the preferences of the users on a particular theme. Based on the sub-modularity of the independent cascade model, the algorithm can guarantee the accuracy and use CELF to improve the efficiency. The part of the experiments shows that the GAT can successfully get the most influential seed nodes on the scientist cooperation network.