Aiming at the computational complexity resulted from the large amounts of support vectors when the support vector machines (SVMs) are used in function estimation, a simplification algorithm is presented to reduce the number of support vectors and simplify applications. By the adaptation of the simplification algorithm, the LS-SVM (least square support vector machine) algorithm can be combined with SMO (sequential minimal optimization) algorithm to achieve good results with high learning efficiency and a few number of support vectors.