Research On Spatial Object Groups Clustering Algorithms Based On Information Entropy

Clustering is an important research problem in data mining, clustering is aimed at partitioning the similar objects of a spatial data set into the same groups,whereas partitioning the dissimilar objects into the different groups and discovering the unexpected, interesting, and useful knowledge. Spatial clustering is an important aspect of spatial data mining, its purpose is to detect clusters or dense area measured by distance from the large data set with multidimensional,two objects from one cluster are similar to each other, whereas two objects from distinct clusters are not. Spatial clustering analysis is a very significant task of data mining.It has been growing concern widely, and is applied to many areas.Now, only the point objects or objects with same feature are considered in spatial clustering algorithms, but the space objects groups' comprehensive features based on the topological relations including inclusion, adjacent and so on are not taken into account. In this paper, the spatial objects groups clustering algorithms based on the spatial relations are studied, some results are as follows:1. Under comprehensively analyzing the features of spatial data, firstly, the relevant definitions of spatial objects groups are given.Spatial data set can be partitioned into several subsets of objects based on spatial relations and each subset can be a spatial objects group.Secondly, the constructing process of spatial objects groups is described clearly through the introduction of spatial data preprocessing algorithm. On this basis, finally, in-depth analysis is made for the clustering of spatial objects groups and lays foundation for the further study of spatial objects groups clustering algorithm.2. An algorithm (named ESOGC) is presented for clustering spatial object groups,in which the objects contained by the spatial objects group are regarded as its features.Through the change of information entropy within a same region, ants determine whether to pick up or drop the current spatial object groups to realize the clustering of spatial object groups. The experimental results show that algorithm ESOGC is effective for spatial data analysis.3. An algorithm (named NSOGC) is presented for clustering spatial object groups, the formula for calculating the similarity between two spatial objects groups is given on the basis of the definition of the information entropy.A new clustering algorithm for spatial objects groups(NSOGC) is producted by applying the similarity calculating method to the traditional spatial clustering algorithms.The experimental results show that NSOGC is effective in spatial objects groups clustering and improving the efficiency of clustering for spatial objects groups.