Place Modelling by Clustering Foursquare POIs in a Place2Vec-Based Approach

Abstract Places are geographic entities defined by specific functions with varying levels of specialization. Points of Interest (POIs) serve as proxies for these functions, with their types indicating distinct roles. This study models places by clustering Foursquare POIs based on their geographic proximity and semantic similarity, where semantic relationships are derived from the Place2Vec model, which generates POI embeddings capturing both spatial and functional contexts. The resulting place clusters are analyzed for functional diversity using entropy, revealing whether they offer specialized or multifaced experiences. This approach provides insights into the spatialsemantic structure of places and their functional complexity.