KM-DBSCAN: Density-Based Clustering of Massive Spatial Data with Keywords

초록

Density-based spatial clustering is one of the most famous algorithms for spatial clustering. Most clustering studies regarding the density-based spatial clustering of applications with noise (DBSCAN) have focused on efficiently finding groups for spatial objects with numeric data or categorical data. In this paper, we propose a novel density-based spatial clustering algorithm called keyword-matched DBSCAN (KM-DBSCAN), which deals with textual data as well as numerical or categorical data. KM-DBSCAN utilizes an IR2-tree as an index structure. To find keyword-matched objects in large datasets, it performs a nearest neighbor search using a branch-and-bound method. While traversing the IR2-tree, KM-DBSCAN effectively prunes unqualified nodes by utilizing both spatial and textual information in nodes. To demonstrate the efficiency of KM-DBSCAN, we conducted extensive experiments in various settings. The experimental results show that KM-DBSCAN is very efficient in terms of computational cost.

키워드