Abstract: ［Objective］ Improper control of metro shield machine attitude deviation will adversely affect the service status of the formed tunnel. Predicting the attitude of shield machine in the construction process is the premise for timely adjusting the attitude, but most of the existing prediction models have problems such as poor interpretability and high data requirements. ［Method］ To increase the interpretability for models, the excavation index SE (specific energy), representing the excavation state of the shield machine in surrounding stratum, is introduced as a characteristic parameter of the model. The shield machine attitude prediction model is established using support vector regression method, which has advantages in small sample learning. K-fold cross validation is used to tune hyperparameters and evaluate the performance and the generalization ability of the prediction model. ［Result & Conclusion］ The integrated model is applied to Chongqing Rail Transit Line 27 engineering case, the goodness-of-fit R2 of the prediction results of the four parameters characterizing shield machine attitude are 0.94, 0.94, 0.90, and 0.87, respectively. The integration of excavation index improves the average prediction accuracy of support vector regression model by 11.96%. Compared to the back propagation neural network model, this integrated model improves prediction accuracy by 6.41%. By introducing characteristic parameters with physical significance, the support vector regression model can more accurately predict the shield machine attitude and provide effective support for real-time shield machine attitude adjustments during the construction process.