Abstract: Measuring sediment transport is both labor-intensive and time-consuming,which necessitates the use of alternative techniques to derive sediment transport information from variables that are more readily measurable.This study aims to leverage machine learning models to accurately predict sediment transport in the Kuye River Basin and to unveil the key hydrological,meteorological,and anthropogenic factors that drive these processes.We employed six machine learning models—Random Forest (RF),Support Vector Machine (SVM),Light Gradient Boosting Machine (LGBM),Extreme Gradient Boosting (XGBoost),Lasso Regression,and K-Nearest Neighbors (KNN)—to predict sediment transport in the Kuye River Basin.Furthermore,we applied the Shapley Additive Explanations (SHAP) approach from cooperative game theory to ascertain the significance and interpretability of various influencing factors on sediment transport prediction.The results show that all six machine learning models performed commendably in predicting sediment transport,with the RF model demonstrating the lowest prediction error and highest efficiency (RMSE=0.51,R²=0.95,MAE=0.38,NSE=0.94).The SHAP analysis highlighted runoff as the most critical hydrological factor and emphasized the significant role of anthropogenic factors such as dam areas,grasslands,forests,and mining areas in mitigating sediment transport,whereas the influence of meteorological factors was comparatively minor.Additionally,the analysis revealed the impact trends of various factors on sediment transport,reflecting the high explanatory power of the models.The findings demonstrate that machine learning models hold significant promise for accurately predicting river sediment transport,particularly in elucidating complex processes.This research provides valuable scientific support for future sediment modeling,soil degradation analysis,and the design of soil and water conservation measures.