Bayesian Network Based Behavior Prediction Model for Intelligent Location Based Services

Abstract

The rapid development in wireless communication and mobile computing brings the booming of intelligent location-based services (LBS), which can actively push location-dependent information to mobile users according to their predefined interests. The successful development and deployment of push-based LBS applications rely heavily on the existence of a spatial publish/subscribe middleware that handles spatial relationship. However, in a traditional publish/subscribe middleware; the current location of a mobile user is the unique criteria to determine whether to notify them. Statistics shows that the accuracy of notification is not satisfied. This paper presents a novel user behavior prediction model (UBPM) for the publish/subscribe system. UBPM is a complementary component of existing publish/subscribe system which is utilized to predict the behavior of a mobile user. This model takes some foregone and real-time user information into consideration that is a prerequisite to predict the future behavior of mobile users. Six important user context-aware information entries which have crucial effects on prediction result are discussed in detail. Furthermore, Bayesian network (BN) and inference in the field of artificial intelligence is introduced to make the prediction more accurate