Abstract: A series of tensile tests were carried out for deformed rare earth magnesium alloy samples with different processes, and a prediction model of mechanical properties of Mg-9Gd-Y-Zn-Zr alloy was put forward with the help of convolutional neural network model training.The results show that, the relative average errors of tensile strength and yield strength of the test set are 3.25% and 4.77% respectively.Three samples of Mg-9Gd-Y-Zn-Zr alloy are randomly selected to verify the predicted mechanical properties of the model again, and the errors are all within 5%, indicating that the convolutional neural network model can predict the mechanical properties with high accuracy.The contribution of fine grain strengthening to the strengthening of the three samples is 68.8 MPa, 88.4 MPa and 98.1 MPa, respectively, and the contribution of dislocation strengthening is 17.62 MPa, 22.36 MPa and 22.06 MPa respectively.