A Cluster-Based Approach for Extracting Redundancy-aware Top-k Co-location Patterns

Spatial co-location pattern mining, which discovers interesting and potentially useful neighboring relationships among spatial features from spatial datasets, plays an important role in spatial data mining. However, as the scale of spatial datasets gradually increases, traditional co-location pattern mining algorithms generate numerous patterns, which may confuse, even mislead the following decision-making of the users. To solve this problem, this paper proposed a cluster-based approach to discover non-redundant top-k co-location patterns, where k is the number of generated patterns. In this proposed approach, first, the spatial data is transformed into transactions to calculate the distance between two co-location patterns, and then the k-means clustering method is adopted to classify the patterns into k clusters, finally, the outputs are generated by selecting one pattern from each cluster. The experimental results on both synthetic and real-world data sets demonstrate that our proposed approach effectively summarizes spatial co-location patterns.