Water inflow (WI) into the tunnel is one of the main geological hazards that can have a significant negative impact on the progress of the tunneling project. A precise prediction of the water inflow into the tunnel, as a significant challenge in rock tunneling, can guarantee project safety and progress. To address this, the current study applied five machine learning (ML) algorithms, including Random Forest (RF), Gradient Boosting (GB), Extreme Gradient Boosting (XGBoost), Categorical Boosting (CatBoost), and Natural Gradient Boosting (NGBoost). Therefore, a dataset including hydraulic conductivity (K), geological strength index (GSI), and water head (H) from 73 sections across six water conveyance tunnels in Iran was collected. The ML algorithms were implemented and then three performance metrics, including the coefficient of determination (R²), normalized root mean square error (NRMSE), and the variance account for (VAF) were employed to evaluate the efficiency of the ML models. As a result, the XGBoost model for testing data showed the greatest level of accuracy and reliability in predicting WI with R², NRMSE, and VAF of 84.9%, 12.9%, and 82%, respectively. Also, based on the results of score analysis and regression error characteristic curve (REC), XGBoost was suggested as the best method for predicting WI in the tunnel. Finally, the water head was found to be the most effective parameter in predicting WI using the Shapley Additive exPlanations (SHAP) and Partial Dependence Plot (PDP) methods.