부동산 거래정보를 이용한 리테일 상권 설정

Establishing appropriate spatial districts is critical in understanding the retail market because market data is surveyed based on them. In general, retail districts have been established from the perspective of usage such as the demand of customers and the supply of retailers. Conversely, this study focused on the transaction of real estate and drew implications by comparing these two approaches. This study analysed statistics of real estate transaction in Seoul using the spatial fuzzy C-means algorithm which is a spatial clustering technique based on machine learning and compared the results with the generally accepted retail districts stemmed from the perspective of usage. The comparison indicates that: First, the transaction-based approach distinguished core and neighbouring retail districts successfully, and the locations of the core retail districts were similar to those of the usage-based approach. Therefore, retail districts could be reliably established using transaction data. Second, despite the similarity in the locations, significant differences were found between the boundaries of the retail districts according to the two approaches. Furthermore, the spatial changes over time, such as the expansion or contraction of retail districts, were identified from the differences. Third, the increase in the prices of retail districts focusing on the transaction was significantly larger than that of retail districts focusing on the usage. This difference indicates that statistical bias in the retail market can occur depending on whether the regionalization is based on usage or transaction.