Title :
Wavelet feature selection for microarray data
Author_Institution :
Shandong Inst. of Light Ind., Jinan
Abstract :
A hybrid method of feature selection based on wavelet analysis and genetic algorithm (GA) is proposed in this study for high dimensional microarray data. A set of orthogonal wavelet approximation coefficients based on wavelet decomposition are extracted to compress the gene profiles and reduce the dimensionality of microarray data. Then genetic algorithm is performed to select the optimized features from approximation coefficients. Linear discriminant analysis (LDA) is employed to evaluate the classification performance. Experiments are performed on four datasets. Our results show that this hybrid method is efficient and robust
Keywords :
cellular biophysics; discrete wavelet transforms; genetic algorithms; genetics; medical diagnostic computing; gene profiles; genetic algorithm; linear discriminant analysis; microarray data; wavelet analysis; wavelet approximation coefficients; wavelet decomposition; wavelet feature selection; DNA; Data mining; Decision trees; Discrete wavelet transforms; Gene expression; Genetic algorithms; Least squares approximation; Linear discriminant analysis; Robustness; Wavelet analysis;
Conference_Titel :
Life Science Systems and Applications Workshop, 2007. LISA 2007. IEEE/NIH
Conference_Location :
Bethesda, MD
Print_ISBN :
978-1-4244-1813-8
Electronic_ISBN :
978-1-4244-1813-8
DOI :
10.1109/LSSA.2007.4400920