High-speed railways are vital infrastructure projects that significantly enhance regional connectivity and collaboration. This study presents an advanced data-driven predictive control methodology to optimize the precision of high-speed train operations. By analyzing operational data across various phases—startup, constant speed, and braking—this study identifies critical patterns and performance variations under different conditions, revealing optimization opportunities. The proposed predictive control system, based on a Random Forest algorithm model, incorporates key processes such as error quantification, predictive control, internal modeling, disturbance compensation, and feedback loops. The internal model utilizes Random Forest algorithms, involving detailed steps like training data analysis, random regression forest development, regression tree calculations, and aggregation of tree estimates to forecast future train states. Simulation and validation results demonstrate that integrating multiple parameters as predictive feature sets in the Random Forest model significantly improves forecast accuracy. Future research could further enhance this machine learning-based control method by incorporating more diverse data sources, thereby improving the accuracy and reliability of train predictions, advancing automation and intelligence in high-speed railways, and promoting sustainability and efficiency.