Abstract: Support Vector Machine (SVM) is a machine learning algorithm based on statistical learning theory, and it has recently been established as a powerful tool for classification and regression problems. This paper introduces the support vector classification and regression algorithms, which are applied to the structure damage identification. With curvature modal parameters as characteristic parameters, a two-step damage location identification approach based on support vector machine is proposed. Firstly, the possible damage location is detected by using support vector classification according to the probability distribution. Then after reconstructing the training set, the precise damage location is identified using support vector regression. The simulation results of the cantilever beam prove that this approach is a promising method for damage diagnosis.