A direct support vector machine for regression(DSVMR) is obtained when appending the square bias of the hyper-plane into the object function in a least square support vector machine for regression(LSSVMR).Compared with LSSVMR,the DSVMR strengthens the convexity of the problem to be solved and only needs to calculate the inversion of a symmetrical positive definite matrix which is similar to a kernel matrix.The Cholesky decomposing and the SMW(Sherman-Morrison-Woodbury) formula are employed to reduce the com...