부산대학교 도서관

Along with the popularization of location-based social networking (LBSN), semantic trajectories, which are trajectories with additional information such as photographs and texts, are increasing, and their utilization is required. We consider frequent pattern extraction as applicable to analysis of semantic trajectories and extraction of regions of interest (ROIs). In this research, we propose SimDBSCAN, which considers both spatial density and similarity of points, by extending DBSCAN, which uses density-based clustering, in order to capture users' interests. Since SimDBSCAN identifies points that are interested in the same object in the neighborhood as ROIs, it is possible to detect not only known ROIs such as tourist sites but also unknown ROIs. In this paper, we explain the algorithm of SimDBSCAN and present the experimental results using photographs collected from Flickr. The experiments show that useful ROIs and patterns can be extracted by the proposed method.