Modeling seed user from short message records

Short messages are an important part of social media network.Besides traditional communications,short messages can also be used for education,entertainment and even business.Nowadays,much work has been devoted to mining useful information and knowledge from short messages via classification and clustering approaches.However,in many situations,it is vital to identify some influential users whose messages will result in a large amount of forwarding,these users are called seed users.In this paper,studies on the forwarding behavior of mobile phone users are carried out first,then,a directed graph-based short message forwarding model is setup,together with a“Reverse PageRank”scoring policy.Finally,the forwarding model and scoring policy have been applied to a real short message record dataset from some telecom operator.By some exploratory data analysis of the dataset,parameters of the forwarding model are determined.Experimental results show that the model and the scoring policy are effective in finding out seed users.