Wavelet based approach to cluster analysis. Application on low dimensional data sets

In this paper, we present a wavelet based approach which tries to automatically find the number of clusters present in a data set, along with their position and statistical properties. The only information supplied to the method is the data set to analyze and a confidence level parameter. Most of the usual methods for cluster analysis and unsupervised classification do not automatically determine the number of clusters present in our data. Thus, the human operator has to supply the method with an a priori number of clusters which the algorithm is expected to find. This fact leads to a difficult interpretation of the resulting clusters. In this paper we also show a practical algorithm to implement this method on low dimensional data sets.