Mining Individual Mobile User Behavior on Location and Interests

With the ubiquitous Internet applications brought by widespread popularity of smart device, an exhaustive understanding of user behavior is becoming essential for Internet Service Providers (ISPs) to implement network management and resource optimization. Existing researches on mobile user behavior principally focus on studying people´s application interests and mobility properties, especially characterizing relationship between the two perspectives. In this paper, distinct from prior work, we propose a new idea to model and predict mobile user behavior after extracting the behavior with strong correlation between browsing interests and location. Initially, improved Apriori algorithm is applied to find the association rule and estimate the strength of correlation between user´s location and application. On that basis, behavior pattern with close correlation between two features is extracted. Subsequently, we use HMM (Hidden Markov Model) to model aforementioned behavior and predict applications used in given location. The effectiveness and accuracy of our method are verified by real data traffic collected from mobile Internet covering 4.51 million people in a large metropolitan area of China over a week.