A Recommendation Agent for Mobile Phone Users Using Bayesian Behavior Prediction

This paper presents a novel agent system which provides a user with useful recommendations of behavior and information based on behavior prediction. The agent understands user’s context from a GPS sensor and a mobile phone, and predicts user’s future behavior based on user’s context. And the prediction of user’s future behavior can be used to provide the user with the suggestion of performing behavior and/or the recommendation of information which are relevant to the predicted user’s behavior. Here, behavior prediction is performed by Dynamic Bayesian Network, which can handle time-series data. The experimental results showed that the proposed agent system was enough to provide convenient recommendation services to users.