A clique-based approach for co-location pattern mining

Abstract

Co-location pattern mining refers to the task of discovering the group of features (geographic object types) whose instances (geographic objects) are frequently located close together in a geometric space. Current approaches on this topic adopt a prevalence threshold (a measure of a user's interest in a pattern) to generate prevalent co-location patterns. However, in practice, it is not easy to specify a suitable prevalence threshold. Thus, users have to repeatedly execute the program to find a suitable prevalence threshold. Besides, the efficiency of these approaches is limited because of the expensive cost of identifying row-instances of co-location patterns. In this paper, we propose a novel clique-based approach for discovering complete and correct prevalent co-location patterns. The proposed approach avoids identifying row-instances of co-location patterns thus making it much easier to find a proper prevalence threshold. First, two efficient schemas are designed to generate complete and correct cliques. Next, these cliques are transformed into a hash structure which is independent of the prevalence threshold. Finally, the prevalence of each co-location pattern is efficiently calculated using the hash structure. The experiments on both real and synthetic datasets show the efficiency and effectiveness of our proposed approaches.