Abstract
The support vector machine (SVM) is a binary classification approach that is both accurate and flexible. It has had significant success, but if too many variables are added, its performance might decrease. The lasso method penalizes least squares regression by adding the absolute values of the coefficients (ℓ1-norm). The structure of this penalty encourages sparse solutions (with many variables coefficients equal to 0). The major goal of group lasso is to construct the lasso, the group formula, in order to find the common elements of groups. The simulation shows that group lasso method outperforms the lasso.
