Data Clustering Using Wave Atom

Abstract

Abstract-Clustering of huge spatial databases is an important issue which tries to track the densely regions in the feature space to be used in data mining, knowledge discovery, or efficient information retrieval. Clustering approach should be efficient and can detect clusters of arbitrary shapes because spatial objects cannot be simply abstracted as isolated points they have different boundary, size, volume, and location. In this paper we use discrete wave atom transformation technique in clustering to achieve more accurate result .By using multi-resolution transformation like wavelet and wave atom we can effectively identify arbitrary shape clusters at different degrees of accuracy. Experimental results on very large data sets show the efficiency and effectiveness of the proposed wave atom bases clustering approach compared to other recent clustering methods. Experimental result shows that we get more accurate result and denoised output than others.