Aiming at the problem that bias of MEMS gyroscope is severely affected by temperature, a temperature compensation method based on Particle Swarm Optimization (PSO) and Support Vector Machine (SVM) is proposed. Firstly, the smoothed gyroscope data is taken as the sample point, and the drift model is constructed by the SVM method based on radial basis kernel function. The data is mapped from low-dimensional space to high-dimensional space for linear fitting to ensure generalization ability. Then, the PSO algorithm is used to optimize the penalty parameters, kernel function parameters and bias parameters of the SVM, which avoids the blindness of artificial parameter choosing and improves the accuracy of the model. Experimental results show that gyro output accuracy is higher after PSO-adjusted SVM compensation. Compared with the least squares method and the BP neural network method, the variance of the gyro output data is reduced by 81.3% and 57% respectively, and the maximum error is reduced by 54.7% and 48.5% respectively.