Towards Effective Top-k Location Recommendation for Business Facility Placement

In the process of selecting locations for establishing new business facilities, location recommendation offers the optimal candidates, which maximizes the number of customers served to bring the maximum profits. In most existing work, only spatial positions of customers are considered, where social relationships and temporary activities, which are significant factors for candidate locations, are ignored. Additionally, current studies fail to take the capacity of service facilities into consideration. To overcome the drawbacks of them, we introduce a novel model MITLR (Multi-characteristic Information based Top-k Location Recommendation) to recommend locations with respect to capacity constraints. The model captures the spatio-temporal behaviors of customers based on historical trajectory and employs social relationships simultaneously, to determine the optimal candidate locations. Subsequently, by taking advantage of feature evaluating and parameter learning, MITLR is implemented through a hybrid B-tree-liked framework called CLTC-forest (tree). Finally, the extensive experiments conducted on real-world datasets demonstrate the better effectiveness of proposed MITLR.