A Clustering and Selection Method using Wavelet Power Spectrum

In this paper, we consider the data mining problem of gene selection based on expression data and propose a method based on wavelet power spectrum. Wavelet power spectrum of the data has been analyzed and on the basis of the observations, a parameter was defined and used to select the features. Our proposed method has been applied on three types of cancer datasets. The results are in comparable with those of earlier works though our method is very simple and easy to and can be used in conjunction with other classification methods. Also, our technique is robust to noise.