Clustering with obstacles for Geographical Data Mining

Abstract

Clustering algorithms typically use the Euclidean distance. However, spatial proximity is dependent on obstacles, caused by related information in other layers of the spatial database. We present a clustering algorithm suitable for large spatial databases with obstacles. The algorithm is free of user-supplied arguments and incorporates global and local variations. The algorithm detects clusters in complex scenarios and successfully supports association analysis between layers. All this occurs within O(n log n+[s + t] log n) expected time, where n is the number of points, s is the number of line segments that determine the obstacles and t is the number of Delaunay edges intersecting the obstacles.