Aiming at the problem of greater measurement error of pressure sensor in continuously variable transmission (CVT), a method of establishing error compensation model using support vector machine (SVM) was presented. Firstly, the basic structure of error compensation model was determined based on the analysis of the factors inducing measurement error. Secondly, training set was constructed by experiment, and the error compensation model was trained with the training data. Finally, the model parameters were optimized by the genetic algorithm during th e training process. The experiment results indicate that the proposed error compensation model improves the linearity of the pressure sensor, and reduces the maximum absolute error from 0.5MPa to 0.15MPa, which greatly improves the performance and measurement accuracy of pressure sensor. 