Abstract: The used SVM kernel function can not approach to any object function in regression.Aiming at the problem，the construction model of Least Square Wavelet Support Vector Machine（LS-WSVM） is proposed based on conditions of the support vector kernel function and wavelet frame theory.Compared with standard SVM and LS-SVM under the same conditions，experimental results show that LS-WSVM has more excellent performance of feature abstraction in regression and its fitting result is more accurate.