Spatial Neighborhood Clustering Based on Data Field

Based on the theory of data field, each sample point in the spatial database radiates its data energy from the sample space to the mother space. This paper studies the use of the data field as a basis for clustering. We put forward a novel method for clustering, which is a kind of natural clustering method called spatial neighborhood clustering. In the data field, the potential center is identical to the cluster center. The key step of the cluster algorithm is to find the potential centers in the grid units of data field. The spatial neighborhood cluster method makes use of the distribution property of the potential value point as the potential center in the data field to discriminate the maximum potential value in a certain neighborhood window. Then the cluster centers can be acquired corresponding to the maximum potential values and the number of cluster centers is automatically amount to that of potential centers.