Clustering Using Wavelet Transformation

This paper introduces and describes an alternative clustering approach based on the discrete wavelet transform (DWT) which satisfies requirements that other clustering methods, like discriminative-based clustering and model-based clustering approaches, do not satisfy. The clustering method has been constructed using wavelet analysis that has the ability of decomposing a data set into different scales. Wavelet algorithm is then used to specify the number of the clusters and quality of the clustering results at each scale. The same algorithm can be generalised for more than one-dimensional data. Some examples about how to use this approach are presented in the paper using different sample sizes and where different kinds of noises are imposed on simulated data. These examples show the successfulness and efficiency of this kind of methodology in detecting clusters under different situations.