Machine learning-based method to predict influential nodes in dynamic social networks

A challenging issue in complex networks is to effectively predict a set of influential nodes. Most previous studies typically ignore the local updates effects on the diffusion process and predict only interactions between nodes. For a good prediction of the influential nodes, each node structure and semantic features can help. In this paper, we propose a novel approach DPIN, a machine learning-based approach, to predict future influential nodes taking into account the structural and semantic characteristics of nodes. We apply this method to predict "hot" papers. Semantic features correspond to "hot" topics detected by the LDA model and structure features correspond to bibliometric features. This approach can help authors make good choices about which topic to target, which article to read, which article to cite and which collaboration to favor. Finally, DPIN experimentations on the DBLP dynamic social network confirm the high quality of predictions.