Cancer Classification Based on the "Fingerprint" of Microarray Data

Recently there is an increasing interest in changing the criterion of tumor classification from morphologic to molecular. In this perspective, the problem can be regarded as a classification problem in machine learning. In this study wavelet analysis is used to extract the features from high dimensional microarray profiles. To make it easier to find the significant genes, we remove the small change contained in the high frequency part based on wavelet decomposition. A set of orthogonal wavelet approximation coefficients is used to compress gene profiles and reduce the dimensionality. Experimental results show that approximation coefficients at 1^st and 2^nd level achieve good performance.