Abstract: In order to predict the load pressure of the sugarcane harvester cutter in the soil cutting process more accurately and quickly, the locomotive speed, soil moisture content, soil density, cutter depth and sugarcane density are used as model inputs, based on regularized twin support. The vector regression machine（ITSVR） model combines the particle swarm optimization algorithm based on genetic algorithm to simulate and predict the cutter load pressure, and compares the simulation results with BP neural network, support vector regression（TSVR）, extreme learning machine（ELM） and The Twin Support Vector Regression（TSVR） model was compared and analyzed, and finally the feasibility of the prediction system to verify ITSVR was established. The results show that the RMSE, MAE and R~2 of the regularized twin support vector regression machine（ITSVR） model are 0.015 4, 0.010 8 and 0.955 8, respectively, and the prediction time of the model is 5.9 ms. The model can load the load more accurately and quickly. The pressure is fitted to the nonlinear curve affecting the parameters; the ITSVR model has better fitting ability and faster prediction speed than other models; when the 5% and 10% relative error are used as the qualified index model, the ITSVR model predicts The correct rates are 66.7% and 100%, respectively, which is a strong proof of the feasibility of the ITSVR model. The research results can provide reference for the subsequent design of automatic cutting control system.