AN IMPROVED CLUSTERING METHOD BASED ON DENSITY AND DIVISION METHOD

Abstract

Combining partitioning and density - based clustering method, an improved clustering method is proposed in this paper on the basis of objective clustering algorithm.  Firstly, the points with greater density which are distant from each other were selected as the initial centers of K-means clustering.  Then, Use K-means was used to roughly determine the elements contained in each class.  Afterwards, the points with the largest density in each class were searched and taken as the centers to re-conduct K-means clustering.  If a class has more than one maximum density points, then the class will have multiple clustering centers, which make the shape of the class not round any longer and facilitate the classification of irregular shapes.  Finally, by the dipole idea of the objective clustering algorithm, the optimal number of clusters was determined.  The improved algorithm proposed in this paper achieved very good results in the clustering tests on random data set and UCI data set.