A New Spatial Fuzzy C-Means for Spatial Clustering

Fuzzy C-means is a widely used clustering algorithm in data mining. Since traditional fuzzy C-means algorithms do not take spatial information into consideration, they often can’t effectively explore geographical data information. So in this paper, we design a Spatial Distance Weighted Fuzzy C-Means algorithm, named as SDWFCM, to deal with this problem. This algorithm can fully use spatial features to assign samples to different clusters, and it only needs to calculate the memberships one time, which reduces the running time greatly compared with other spatial fuzzy C-means algorithms. In addition, we also propose two new criteria, named as DESC and PESC, for evaluating spatial clustering results by measuring spatial and regular information separately. The experiments are carried out based on real petroleum geology data and artificial data, and the results show that SDWFCM can achieve better performance compared with traditional clustering method, and our spatial cluster indices can provide the assessment of clusters by taking spatial structure into consideration effectively. Key-Words: spatial clustering, fuzzy c-means, evaluation criteria.