Sparse supervised principal component analysis (SSPCA) for dimension reduction and variable selection

Sara Sharifzadeh, Ali Ghodsi, Line H. Clemmensen, Bjarne K. Ersbøll

Abstract: Highlights •A sparse supervised PCA method (SSPCA) is proposed.•It computes sparse Eigen vectors with maximum dependency to the response vector.•SSPCA can be used for data sets with linear as well as non-linear behaviour.•The sparse Eigen vectors can be used for feature selection and has been tested on spectral signals/images as well as microarray data sets.•SSPCA is compared with PCA, PMD-based SPCA, Supervised PCA and SPLS.