Abstract: The mathematical regression model of standard least squares support vector machine(LS-SVM) is presented and a multi-kernel least squares support vector machine(MLS-SVM) algorithm is proposed to enhance the regression accuracy of non-flat functions.The hierarchical clustering method is applied to deal with the problem that the solution of MLS-SVM is lack of sparseness.Partial least squares regression(PLSR) method is adopted to realize robust regression of MLS-SVM.A simulation example is given to validate the effectiveness of the presented method.