Smart Store Understanding Consumer's Preference through Behavior Logs

This paper presents a smart store that estimates a preference of consumers concerning products from their behaviors. This paper proposes a method, which is a passive observation and an active observation, to observe two behaviors, direct behaviors and indirect behaviors. The passive observation is a method to observe direct behaviors of customers towards real products through ambient sensors. The active observation is a method to observe indirect behaviors of customers towards information of products through ambient displays. This study explains a purchase experiment using a prototype smart store that has installed the ambient shelves and displays. This study estimates the favorite clothes from their direct and indirect behavior using the smart store. The result of estimation of preference shows that accuracy rate is 87% by leaveone-out cross-validation.