Chain store placement recommendation with multi-source urban data

With the rapid development of chain stores, how to recommend the placement of new stores becomes a challenge. Existing work mainly focus on using feature-based models to recommend places for stores. Different from existing works, we focus on the scale-based store placement problem, which cannot recommend places for a store, but also suggest the right scale for the store. Firstly, we extract geospatial features, business features and scale-relevant features to characterize the urban regions. Secondly, we present regression and classification methods for dealing with two situations: recommending places with a given scale, and recommending scale given the place. Experiments over real world datasets prove the effectiveness of our method.