Wavelet high-frequency coefficients for feature extraction of gene microarray data

In the paper,we use the wavelet analysis theory and the support vector machine theory to build a model which can classify the prostate cancer microarray data into cancer and normal classes.We mainly research the wavelet high-frequency coefficients for feature extraction of prostate cancer gene microarray data in contrast to the low coefficients.We extract haar wavelet high-frequency coefficients at level 3 and feed the high-frequency coefficients to the classification.The correct classification rate is 93.31%.We extract db1 wavelet low-frequency coefficients at level 4 and feed the low-frequency coefficients to the classification.The correct classification rate is 93.53%.The wavelet low-frequency coefficients for feature extraction are better than high-frequency coefficients.The classification modle is very stable.