Mining Co-location Patterns with Spatial Distribution Characteristics

Spatial co-location patterns represent the subsets of Boolean spatial features, and the instances of the pattern are frequently located together in a geographic space. Most existing co-location pattern mining methods mainly focus on whether spatial feature instances are frequently located together. However, that the occurrence of neighbor relationships is in the whole space or local area is not considered. In this paper, a new measurement using an evenness coefficient of the feature distribution is introduced, and a novel algorithm for co-location pattern mining is proposed, which takes into account the prevalence of the spatial feature and the spatial distribution characteristics of feature instances. Furthermore, some key techniques are presented, including region partition and count of row instances in this algorithm. The experimental evaluation with both synthetic data sets and a real world data set shows that the algorithm can discover prevalent and evenly distributional co-location patterns, and the number of the result set is effectively reduced compare to the traditional mined results.