A framework for mining spatial high utility co-location patterns

Abstract

A spatial co-location pattern is a group of spatial features, whose instances frequently appear in the same region. In association rule mining, a recent effort has been to consider utility as a new measure of interests, by considering the different values of individual items as utilities. In this paper, we incorporate utility into the spatial pattern mining through the concept of pattern utility, and a general framework for spatial high utility co-location patterns mining is defined. Furthermore, we define the concept of extended pattern utility ratio and present an Extended Pruning Algorithm (EPA) to prune down the number of candidates and can obtain the complete set of spatial high utility co-location patterns. Using synthetic and real-world data sets, substantial experiments show that EPA is effectively and efficiently identifies high utility patterns from spatial datasets.