The bearing is an important part of rotating machinery, and the quantitative diagnosis of bearing faults is of great significance in maintaining behavior. Considering the nonlinear relationship between the bearing feature parameters and fault size, a support vector regression (SVR) machine is introduced, and the quantitative diagnosis strategy and procedure based on the SVR is proposed. Based on the extracted feature vectors of the bearing under different fault sizes, the SVR quantitative diagnosis model is constructed and applied to the quantitative identification of the bearing fault. The results verify the effectiveness and improved capability of the proposed SVR method for the quantitative identification of the bearing fault, compared with the artificial neural network based strategy.