In order to efficiently and accurately predict the crack propagation life of friction stir welding （FSW） welded structures， a fatigue crack propagation life prediction method for 2195 aluminum-lithium alloy FSW joints based on support vector regression （SVR） is proposed. Firstly， the Paris crack propagation model constants and the crack tip stress intensity factor dataset are obtained by the fatigue crack growth test and the finite element simulation. Then a stress intensity factor prediction model is established based on the SVR model， and the hyperparameters of the SVR model are optimized by the particle swarm optimization （PSO） algorithm. Finally， the crack propagation life is predicted based on the stress intensity factor prediction model and the Paris model. The results show that the optimized PSO-SVR model can effectively predict the crack tip stress intensity factor， and the determination coefficient R² on the test set can reach 0.999 5， which is higher than 0.954 of the unoptimized SVR model. The prediction results of crack propagation life by the proposed method are compared with those by the finite element simulation. The maximum error is less than 5%， verifying the accuracy of the method.