Mining co-location patterns with dominant features

Abstract

The spatial co-location pattern mining discovers the subsets of spatial features which are located together frequently in geography. Most of the studies in this field use prevalence to measure a co-location pattern’s popularity, namely the frequencies of a spatial feature set participating in a spatial database. However, in some cases, users are not only interested in identifying the prevalence of a feature set, but also the features playing the dominant role in a pattern. In this paper, we focus on mining dominant-feature co-location pattern (DFCP). We firstly propose a new measure, namely disparity, to measure the disparity of features in a pattern. Secondly, we formulate the DFCP mining problem to determine DFCP and extract dominant features. Thirdly, an efficient algorithm is proposed for mining DFCP. Finally, we offer an experimental evaluation of the proposed algorithms on both real data sets and synthetic data sets in terms of efficiency, mining results and significance. The results show that our method can effectively discover DFCPs.