Unveiling spatial semantics of urban POIs for tourism-oriented location-aware business recommendations

The choice of launching a commercial activity relies on a variety of factors, which can deeply influence the type of business and its physical location, especially when the activity involves a tourism-oriented target. Whereas traditional research works primarily focus on the economic and management-related features affecting a potential profitable enterprise, the paradigm of ‘optimal location for an optimal business’ has mostly been neglected. Our study aims to provide an innovative solution for exploring spatial semantics of business-related POIs, employing a spatially aware unsupervised embedding strategy built on artificial neural networks. The purpose is to generate dense vectors of meaning that are able to unveil recurrent geographic relations among business types, consequently identifying the most favorable areas for opening a specific enterprise or recommending the most promising activities for a selected location. The experiments involve a real-world integrated dataset comprising various POI subtypes, including restaurants, hotels, tourist attractions and further venues.