Research On Mobile App Usage Prediction Based On User Behavior Pattern

With the development of wireless network technology and the popularity of mobile devices, more and more mobile applications have been used in people’s daily life. With the explosive growth of mobile device usage, helping the users obtain the accurate service efficiently has become a hot issue in mobile service field. By mining the user behavior pattern to forecast the user’s next target application, the mobile system can prepare the running context in advance, which will benefit to the reduce of user operation time and error operation, thus helping improve the user experience. Obviously, due to the complexity and polytrope of the mobile context, extraction of mobile user behavior patterns is more complex than those in traditional environment.The main work of this paper is as follows:Firstly, based on real users’ mobile application use data, we analyzed the mobile users’attributes and behavior patterns respectively. After abstracting the data into time series and a series of analysis, we got some change characteristics of mobile user patterns, based on which we proposed three kinds of prediction model step by step.Secondly, as for the problem of mobile users’behavior prediction, the paper first proposed two kinds of session-time-based algorithm to get the mobile user behavior sequences. Then mining the history data in our mining algorithm, which is based on "PrefixSpan" algorithm. Finally, we raised up two kinds of predication model based on user behavior patterns.Finally, in order to prove the feasibility of the models described above, we designed two experiments based on real usage data, combined with a traditional statistical model as a baseline comparison test. Through our experimental test, prediction models based on user behavior patterns is showed to have a better performance in terms of prediction hit radio, these models provide a feasible schema for improving use experience in mobile environment.