Interval Support Vector Machine In Regression Analysis

Support vector machines (SVMs) have been widely applied in regression analysis. In this paper, the application of SVM in regression for interval samples is proposed. The standard support vector regression (SVR), is a quadratic optimization problem that is formulated according to the form of training samples and optimal hyperplane is obtained. In real world, the parameters are seldom known and usually are estimated. In this paper we propose an interval support vector regression (ISVR) problem which the training samples are interval values. Using duality theorem and applying variable transformation theorem the problem is solved and two hyperplanes correspond to the upper bound and the lower bound of solution set is obtained. Efficiency of our approach is confirmed by a numerical example. Keywords: Support vector machine, Regression analysis, Interval quadratic optimization problem. 1. Introduction Recently, a novel machine learning technique, is called SVM, drawn much attention in the fields of pattern classification and regression forecasting. SVM was first introduced by Vapnik in 1995 [2]. SVM is a kind of classifier’s studing method on statistic study theory.