The work aims to improve the performance of surface acoustic wave resonator (SAWR), and manufacture a high-performance surface acoustic wave temperature sensor. Based on the FEM/BEM theory, an accurate simulation optimization model of SAW temperature sensor was established, and the Euler angle of the sensitive substrate was optimized in large steps based on this model. At the same time, the Euler angle of the sensitive substrate was optimized quickly in small steps with the polynomial regression model in combination with the simulation data. The optimization design method combining FEM/BEM simulation model and machine learning proposed in this paper could not only realize the accurate simulation of SAWR, but also greatly improved the optimization efficiency. The relative error of the optimized result and the actual device's center frequency was 0.4%, and the relative error of the Q value was 1.2%. Compared with the pure FEM/BEM method, its speed of single cutting calculation was increased by more than 2000 times. The designed optimization system can be used to quickly optimize the design of the resonator's sensitive substrate cutting, which can shorten the development cycle of high-performance SAW temperature sensors.