Abstract: Full-face hard rock Tunnel Boring Machines (TBM) often encounter weak, loose and unstable surrounding rock conditions during tunnel construction, which will easily lead to rock collapse and cause the cutter head or shield to be stuck.Accurately and timely predicting the surrounding rock conditions at TBM face is of great significance for adjusting the excavation and support parameters in a timely manner and ensuring construction safety.This study took the Water Supply Project in the Central City of Jilin Province as an example and proposed a TBM tunnel surrounding rock grade prediction method based on the Sparrow Search Algorithm (SSA) and RF, XGBoost, SVM, LSTM four machine learning methods.The research findings indicate: SSA can improve the prediction accuracy of machine learning models to varying degrees.Through the comparison of model performance indicators, the prediction performance of SSA-RF and SSA-SVM optimization models were better than that of SSA-XGBoost and SSA-LSTM models.SSA-RF and SSA-SVM were selected as base models and combined into an integrated model SSA-RF-SVM through weight optimization.Through model performance analysis, SSA-RF-SVM performed best in the test set.Comparing the performance of all established models, the prediction accuracy for surrounding rock classification decreased in SSA-RF-SVM, SSA-RF, RF, SSA-SVM, SSA-XGBoost, SSA-LSTM, XGBoost, SVM, LSTM order.