Mining Customer Preference in Physical Stores From Interaction Behavior

An improved understanding of customer preference is crucial for successful business in physical stores. Online stores are capable learning customer preference from the click logs and transaction records, while retailers with physical store still lack effective methods to in-depth understand customer preference. Fortunately, user-generated data from mobile devices and social media are providing rich information to uncover customer preference. In this paper, we present a novel approach to mine customer preference in physical stores from their interaction behaviors. To demonstrate the utility of the proposed model, we conduct a store-type recommendation model for physical stores by jointly considering the learned customer preference and temporal influence. We have performed a comprehensive experiment evaluation on two real-world data sets, which are collected by more than 120 000 customers during 12 months from two urban shopping malls. Experimental results show the superiority of the proposed model not only in recommending interesting stores for customer, but also help retailers better understand customer preference.