<p>Rail transport systems developed into high-speed and technologically advanced systems that focus on the delivery of efficient and reliable services. Poor punctuality and frequent train delays erode passenger confidence and satisfaction considerably, which is a major challenge to railway operators. Delays arise from causes as varied as accidents, equipment failure, operational inefficiencies, maintenance, construction activities, inclement weather conditions, and passenger-related incidents. This study to overcome these challenges develops an optimized machine learning model to estimate delays and ensure reduced operational expenditure and improved customer satisfaction. These predictive models use train delay estimation to smooth all operational processes, which offers huge improvements in passenger experience and rail transportation reliability. The proposed research focuses on modelling rail transport systems using regression models such as Random Forest, Decision Tree, K-Nearest Neighbors, Bagging, Extra Trees (ET), and Gradient Boosting to estimate train delays. The performance optimization of these models was carried out using an Enhanced Chimp-Harris Hawks Optimization Algorithm. The performance analysis of the models indicates that the ET model with the highest R² score of 0.95, and the lowest RMSE of 45.93 emerge as the most reliable approach for forecasting train delays. The sensitivity analysis of the ET model introduces the distance between stations variable as the most important feature in a variation of train delay value, around 95%, referring to the main factor causing delays in the rail transport system. These results provide an in-depth overview of effective factors on train delay, which equips the rail operator with an effective toolkit that improves schedule accuracy and minimizes delays.</p>