Retail District Clustering in Seoul Utilizing Machine Learning Method Based on Large- and Small-Sized Retail Stores’ Sales and Location Data

Abstract

Although retail industy including the fashion sector encompasses a range of stores sizes from large to small, existing literature offers inconsistent views about the relationship between large-sized and small-sized retail stores. This study aimed to cluster retail districts in Seoul based on the location and sales data of large-sized and small-sized retail stores, employing the Self-Organizing Map (SOM) as a kind of machine learning approaches. The results revealed major retail clusters by sales, location, and store sizes, and the descriptive characteristics of these clusters, such as, land price, living population, and the number of subway lines. These clusters were then further divided into three categories: mutually dominant clusters, large-sized store dominant clusters, and small-sized store dominant clusters. The results show which retail districs are more advantageous for one of the store sizes or for both, thus providing an empirical, data-driven insight to develop optimal location strategies for retailers and policy-makers.