Exploiting User Context and Network Information for Mobile Application Usage Prediction

The explosive increasing mobile Applications (Apps) have been attracting researchers and developers to investigate user preferences on various mobile Apps. Understanding mobile Apps usage pattern of end users will help to improve the quality of mobile Apps and meanwhile enhance the quality of experience of users. For instance, if the mobile device knows the App the user will launch, it can pre-load the App into memory and also put the quick launch icon on the home screen to speed up the App usage. In this paper, we collect mobile data from over 10,000 users and study the mobile Apps usage pattern from user perspective using the collected data on mobile devices, including time, location, last used App, network type, network speed and etc. We then use Naive Bayes and linear model to propose mobile App usage prediction method for the prediction of the next launched App by individuals. The result shows that the proposed App usage prediction method can reach 60% accuracy to predict the mobile App that will be launched by users.