A Method For Partioning Even Clusters Based On Grids

With the development of information technology, we can get more and more information in daily life. In this situation, how to select useful information for us from a lot of information becomes a major research projects for many scholars. To solve this problem, we developed data mining and it is becoming more and more important. Clustering, as a main technology in data mining, plays an important role and develops a lot in recently years, especially the clustering method based on grids become a main research project because of its advantage of fast computational speed.However, all the existing kinds of clustering method cannot get even clusters. Therefore, this paper puts forward a new method for partitioning even clusters based on grids.In this article, concepts, techniques and algorithms about data mining will be discussed firstly. Then we discuss the beginning, concepts, technology of clustering. We divide the existing clustering methods into two main catalogs and five steps. Besides, we point out the problems that we should pay much attention to in every step, so we can get more reasonable results when we use clustering technology.Secondly, the traditional clustering algorithms are discussed in detail, including every kind of methods'algorithm thought, algorithm flow, advantages and disadvantages. And we compare these kinds of methods and make a form, so we can choose the suitable method when we use clustering technology.Thirdly, the article points out that the existing clustering methods just put the "similar" points into a cluster and the number of points in every cluster is different. However, sometimes, we want to get even clusters, that is, the number of data in every cluster is equal. Obviously, the existing clustering method cannot meet the user's constraints. So in this article we give a solution of this problem. We propose a new method for partioning even clusters based on grids. By bisecting and merging the grids, this algorithm will gather certain grids into one cluster. In this way, the result of this method not only ensures that the points with similarity are divided into one cluster but also ensures the number of points in every cluster is equal. This algorithm has two advantages:one is that it has fast calculation of Grid Computing, the other is it overcomes the obstacle that existing clustering methods cannot get even clusters. We finished this algorithm using MATLAB and the simulations verify the validity of this method.At last, we summarize this paper. We prospect the application of this new algorithm of clustering method and point out how to improve this algorithm.