Estimating Potential Customers Anywhere and Anytime Based on Location-Based Social Networks

Acquiring the knowledge about the volume of customers for places and time of interest has several benefits such as determining the locations of new retail stores and planning advertising strategies. This paper aims to estimate the number of potential customers of arbitrary query locations and any time of interest in modern urban areas. Our idea is to consider existing established stores as a kind of sensors because the near-by human activities of the retail stores characterize the geographical properties, mobility patterns, and social behaviors of the target customers. To tackle the task based on store sensors, we develop a method called Potential Customer Estimator (PCE), which models the spatial and temporal correlation between existing stores and query locations using geographical, mobility, and features on location-based social networks. Experiments conducted on NYC Foursquare and Gowalla data, with three popular retail stores, Starbucks, McDonald’s, and Dunkin’ Donuts exhibit superior results over state-of-the-art approaches.