HDBSCAN 클러스터링 방법을 이용한 도시 관심 지역 추출에 관한 연구

초록

Since 2020, the COVID-19 pandemic has caused a variety of social and economic problems worldwide, leading to a rapid increase in online activities and the use of social network services (SNS). SNS data can provide a large amount of low-cost data with diverse information that can be utilized for urban planning and management. In response, it is possible to explain the change of spatial patterns in various ways using advanced artificial intelligence technology. While K-means and DBSCAN algorithms have been widely used for clustering, the applications of HDBSCAN in domestic research has been rarely performed. Therefore, this study aims to detect the city’s interest areas using HDBSCAN with Flickr data from Seoul in 2019 and 2020. In this study, HDBSCAN is used to cluster Flickr posts and extract urban interest areas, and the clustering level for each cluster is evaluated using silhouette scores. As a result, clusters are extracted along the major points in downtown Seoul and suburbs, and the cluster level was found to be statistically significant based on the silhouette score.

키워드