Machine Learning Techniques for SVM Classification and Regression Loss Function

Abstract

Support vector machines (SVM) are a group of supervised learning methods that can be applied to classification or regression. SVM regression is used to predict various physical, chemical, or biological properties). A presentation of linear SVM followed by its extension to nonlinear SVM and SVM regression is then provided to give the basic mathematical details. SVMs were developed to solve the classification problem, but recently they have been extended to the domain of regression problems. In this paper gives the idea of SVM regression is based on the computation of a linear regression function in a high dimensional feature space where the input data are mapped via a nonlinear function. SVR attempts to minimize the generalization error bound so as to achieve generalized performance instead of minimizing the observed training error.Statistical Learning Theory has provided a very effective framework for classification and regression tasks involving features. Support Vector Machines (SVM) are directly derived and they work by solving a constrained quadratic problem where the convex objective function for minimization is given by the combination of a loss function with a regularization term (the norm of the weights).