Mining mobile application sequential patterns for usage prediction

In recent years, researches on smart phone services have received a lot of attention in both of the industry and academia due to a wide range of potential applications. Among them, one of popular topics is the mining and prediction of mobile application usage behaviors. In this paper, we propose a location-based approach to predict the mobile application usage behaviors. In this approach, we first discover the stay locations of the GPS movement data to obtain the mobile application usage database. Then, we consider both of the physical location moving paths and virtual application usage paths of users to mine the Mobile Application Sequential Patterns (MASPs) by the Mobile Application Sequential Pattern Mine (MASP-Mine) algorithm. Furthermore, a prediction strategy is designed to predict the next mobile application usage behaviors based on the MASPs. To our best knowledge, this is the first work on mining and prediction of mobile application usage behaviors with considerations of physical location moving paths and virtual application usage paths simultaneously. Through experimental evaluation under various condition settings by a real mobile application usage data, the proposed approach is shown to deliver excellent performance in terms of precision and recall.