Activity Prediction Using Truval Method In Mobile Social Network

Now a days current trend of online social network is turning towards mobile. Mobile social networks directly reflect our real social life, therefore are an important source to analyze, understand the underlying dynamics of human behaviors (activities). In this report, we are going to study the problems of activity prediction in mobile social networks. We propose a series of observations in two real mobile social network and then propose a method, Truval in activity prediction based on a dynamic factor-graph model for modeling and predicting user activities. An equivalent algorithm based on mean field is presented to efficiently learn the proposed method. We are going to deploy a actual system to collect users day to day activity, behaviors and validate the proposed method on two combines mobile datasets. Shows that the proposed ACTPred model using Truval method can achieve better performance than baseline methods.