Discovering Prevalent Co-location Patterns Without Collecting Co-location Instances

Discovering prevalent co-location patterns (PCPs) is a process of finding a set of spatial features in which their instances frequently occur in close geographic proximity to each other. Most of the existing algorithms collect co-location instances to evaluate the prevalence of spatial co-location patterns, that is if the participation index (a prevalence measure) of a pattern is not smaller than a minimum prevalence threshold, the pattern is a PCP. However, collecting co-location instances is the most expensive step in these algorithms. In addition, if users change the minimum prevalence threshold, they have to re-collect all co-location instances for obtaining new results. In this paper, we propose a new prevalent co-location pattern mining framework that does not need to collect co-location instances of patterns. First, under a distance threshold, all cliques of an input dataset are enumerated. Then, a co-location hashmap structure is designed to compact all these cliques. Finally, participation indexes of patterns are efficiently calculated by the co-location hashmap structure. To demonstrate the performance of the proposed framework, a set of comparisons with the previous algorithm which is based on collecting co-location instances on both synthetic and real datasets is made. The comparison results indicate that the proposed framework shows better performance. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.