Sample-space-based feature extraction and class preserving projection for gene expression data

摘要

In order to overcome the problems of high computational complexity and serious matrix singularity for feature extraction using Principal Component Analysis (PCA) and Fisher's Linear Discrinimant Analysis (LDA) in highdimensional data, sample-space-based feature extraction is presented, which transforms the computation procedure of feature extraction from gene space to sample space by representing the optimal transformation vector with the weighted sum of samples. The technique is used in the implementation of PCA, LDA, Class Preserving Projection (CPP) which is a new method for discriminant feature extraction proposed, and the experimental results on gene expression data demonstrate the effectiveness of the method.