Understanding metro station areas' functional characteristics via embedding representation: A case study of shanghai

As crucial transportation hubs for urban travel, metro stations catalyze the transformation of their surrounding areas into highly prominent locations where many activities converge. Uncovering the functional attributes of station areas holds immense significance in comprehending citizens' activity demands, thereby offering valuable insights for regional development and planning in proximity to metro stations. This study introduces a framework that improves the process of accurately representing station areas. On the basis of the semantic vectors of point of interests (POI) categories trained by the GloVe model, the partition smooth inverse frequency (P-SIF) model and affinity propagation (AP) are employed to generate the embedding representations of station areas and categorize. Finally, we classify the station areas into 9 functional groups: and analyse the spatial distribution characteristics of each group. It is found that most of the station areas in Shanghai show the characteristics of mixed type, in which the characteristics of residential type and commercial type are obvious. In terms of spatial, the stations with commercial characteristics are mainly distributed in the central area of the city, while those with residential and working characteristics are scattered.