Feature selection based on step-wise extraction of partial least square principal components

Feature selection techniques are widely used in biological informatics.In this paper,partial least-squares(PLS) is repeatedly used for extracting principal components.By selecting the genes which have larger weights in principal components,PLS is introduced for feature selection.This method is used in tumor gene expression profiling data,and the feature genes selected is used for classification.Precision of 92.5% can be achieved when using 8 feature genes.