Enhancing Mobile Recommender Systems with Activity Inference

Abstract

Today’s mobile leisure guide systems give their users unprecedented help in finding places of interest. However, the process still requires significant user interaction, specifying preferences and navigating through lists. We argue that this interaction, while effective when used properly, is an obstacle for users trying to learn these systems. Furthermore, ad-placement systems, which functionally resemble recommender systems, must rely on other sources of information to automatically infer the users preferences and interests. In this paper, we describe how automatically infer a users high-level activity to better support recommendations. Activity is determined by interpreting a combination of current sensor data, models generated from historical sensor data, and priors from a large time-use study. We present an initial user study that shows prediction accuracy improvements from 62% to over 77%, and a discussion of the challenges of integrating activity representations into a user model.