The Research Of Topic-Based Influence Maximization Problem In Social Netwo Rks

Recently, with the increasing popularity of Social Networking Services, the problem of finding top-k most influential nodes in social networks becomes more and more important. There are many studies on this problem, but they don't take account of user preferences, which greatly affects the accuracy of results.We propose a new algorithmic solution to the topic-based influence maximization problem. A two-stage mining algorithm (GAUP) has been proposed for mining most influential nodes on a specific topic. In the first stage, GAUP uses Vector-Space Model or Latent Semantic Indexing, to determine user preferences on a topic. Then in the second stage, based on a diffusion model taking user preferences into account, GAUP adopts a greedy algorithm to find top-K nodes in the network. Compared with the previous work, GAUP achieves higher accuracy when given a topic. Using an analysis framework of submodular property, GAUP can obtain an approximate solution that is provably within 63% of the optimal and can use CELF-optimization to improve algorithm's efficiency .Our evaluation on DBLP shows that GAUP algorithm can successfully mine top-k influential nodes for a given topic. Finally, we have used GAUP for finding domain experts.