MCR SVM classifier with group sparsity

Classification and dimensionality reduction of high-dimensional data are two important topics in bioinformatics, data mining and machine learning. We propose a novel sparse minimax concave ridge support vector machine (MCR SVM) classifier that simultaneously performs classification and dimensionality reduction. The MCR SVM classifier proposed in this study combines the advantages of the unbiasedness of the estimators of the SCAD SVM and the ability of feature group selection of HHSVM to overcome the disadvantages. We also provide a theoretical justification for the group sparsity of the selected features. The experiments on artificial highly correlated data and high-dimensional real-world data with a small sample size show that the MCR SVM classifier is a attractive technique of classification and dimensionality reduction and its performance is better than the other sparse SVMs.