Inferring Point-of-Interest Relationship for Strategic Group Discovery Guided by User Demands

Identifying strategic groups is crucial for businesses to understand the local competitive landscape and develop successful strategies. With the advancements in location-based services and mobile edge computing, Points of Interest (POI) data and customer visiting data are used to infer competitive relationships. However, existing methods only detect whether competitive relationships exist between pairwise POIs, disregarding the multifaceted nature of business strategies such as price, target consumer, and site selection, which fails to group businesses with similar strategies. To address this issue, we propose EAGLE, a strategic group discovery framework that models business strategies by exploring groupwise POI relationships guided by user demands. Our framework constructs an urban hypergraph that captures various dimensions of high-order competitive relationships among POIs, brands, city regions, and products/services. Using a query encoder, we incorporate user demands into the urban hypergraph's vector space, identifying strategic groups through hypergraph learning. We design a multi-hop attention mechanism to mitigate the over-smoothing problem and introduce an inductive learning method with a strategy-aware sampling technique to handle dynamic POI data effectively. Our evaluation results demonstrate that EAGLE outperforms state-of-the-art methods in identifying POI strategic groups that match users' diverse search demands.