A Sequential Recommendation for Mobile Apps: What Will User Click Next App?

Abstract

With the advances in smartphones users install abundant apps to facilitate their daily lives. Both users and related developers have increasing requirements to understand the mobile App usage pattern, for individual and commercial use. Respectively, personalized App recommendation methods and systems have emerged as a novel attractive topic that can demonstrate the human App usage behavior. The mobile Apps recommendation can serve as a cornerstone for a variety of intelligent services, such as fast-launching UIs, intelligent user-phone interactions, and battery management of cellphones. In this paper, we develop a novel App recommendation framework combining the historical App usage data with the sequence of recently-used Apps. Specifically, our framework is an extension of the user-based collaborative filtering technique, where the set of nearest neighbors is employed for training the prediction model. However, our prediction scheme is constructed on the temporal sequential data and is modeled by using the chain-augmented Naive Bayes model. Experiments with a real mobile Internet record dataset demonstrate that the accuracy of our framework outperforms several baseline App recommendation approaches.