Journal of the Korean Geographical Society ISSN:1225-6633(Print) 대한지리학회지

Regions have their own unique characteristics, yet it is not easy to effectively take account of such features when making policy to tackle various social and public issues. For that, it is necessary to establish and implement policies that are consistent with regional conditions. However, suggesting regional policy through cluster analyses showed limitations in selecting a cluster analysis method for deriving spatial characteristics, the amount and quality of data used, and linking cluster analysis results with policies that should be recommended. In this study, the Self-Organising Map (SOM), one of prominent machine learning techniques, was utilised to conduct a cluster analysis in Seoul, and the process of regional policy recommendations was evaluated. As a result of the cluster analysis, Seoul was divided into eight clusters representing unique characteristics. Based on this, policies were derived on a cluster basis, considering both the spatial context and characteristics of the clusters. Through the analysis process with SOM, which exploits big data, we demonstrated that it can provide more objective evidence for making regional policy decisions. However, it was also found that variables that did not exhibit noticeable characteristics were likely to be ignored when using SOM for performing such cluster analysis, and for making policy suggestions. It was anticipated that the results of this study could be used as a valuable reference when planning and establishing policies by taking account of regional characteristics.