The Research Of Influence Maximization In Academic Network

With the rapid growth of network papers, how to choose k papers which are the most influential nodes among a large number of academic documents has become a hot issue in the field of data mining. Therefore, it is of great theoretical and practical significance to study the influence maximization problem based on academic network. And it has achieved many research results, mainly including two aspects what are about the improvement of propagation model and the optimization of influence maximization algorithm. However, the research about the propagation model in the past only considers the reference relationship between papers ignoring the link strength between them. And traditional methods only focus on the impact strength of a node itself ignoring the influence of its inactive neighbor node for the research of influence maximization algorithm.Aiming at the problems in the above methods, we study the influence maximization problem in academic network from the following two aspects:(1) The academic network propagation model based on the quality of the paper is improved. The correlation degree of theme for papers those have reference relation is measured by reference correlation degree of papers, and the algorithm of quality measurement based on PageRank is improved. We take into account the influence of a paper who owns high quality on diffusion, and the academic network propagation model is established on the basis of linear threshold model.(2) Propagation algorithm on the basis of the division of the topic network and the potential impact of the network nodes is provided. This method provides the way how the potential impact strength of nodes is measured by the effect that neighbor nodes contribute to the spread when we select seed node in the heuristic stage. And through combining with the division of the topic network and the greedy algorithm, we get the propagation algorithm of the influence maximization problem in academic network and analysis its time complexity.Through the experiment and analysis in ACM data academic network, we can see the effectiveness and validity of the model and the algorithm proposed in this paper.