Discovering Co-location Patterns in Datasets with Extended Spatial Objects

Speaker: Iris ZHANG

Abstract

Co-location patterns are subsets of spatial features that co-locate frequently. Previous research focused on finding co-location patterns of point features. In real applications, there are multiple types of objects, such as line-strings and polygons. The co-location of these extended spatial objects is like “highways in large metropolitan area often have frontage roads nearby”. In this talk, a buffer-based definition of neighborhoods will be introduced. Furthermore, two pruning approach, which are prevalence-based pruning and geometric filter-and-refine, will be described and compared. Experimental evaluation with a real data set, which is digital road map of the Minneapolis and St. Paul metropolitan area, shows that the geometric filter-and-refine approach can prune a lot of features which cannot form co-location patterns so that improve the performance of mining algorithm significantly.