Feature Mapping the Seoul Metro Station Areas Based on a Self-Organizing Map

Land in the vicinity of railway stations in the Seoul metropolitan area has been spotlighted as a target for redevelopment in accordance with the principles of transit-oriented development (TOD). In order to understand the nature of station areas as a whole, it is required to identify their current status with respect to their built environments, demographic characteristics, socioeconomic status, and transport aspects. Most of the previous studies that have focused on characterizing urban areas assumed that clearly separable clustering is possible and tried to find a robust methodology for that. Many researchers have used either supervised classifiers requiring a pre-classified dataset for training, or K-means-like unsupervised classifiers demanding a predetermined number of clusters. The present study focused on the fact that it was hard to find such a clear separation in station areas in Seoul. A more abstract technology was necessary to position the current status of each station area and to find how different station areas are from one another. A robust unsupervised classifier, called the self-organizing map (SOM), was employed to investigate the similarities and differences among station areas in Seoul. The SOM results revealed many informative findings for policy development without any classification.