Prediction of train delays at multi-line railway stations using machine learning

Increasing demand for railway transportation makes it more difficult to predict train delays and and reduces operational reliability by intensifying route conflicts at multi-line stations. In this study, the aim is to predict train arrival delays using real-time data collected over the course of one week from Prague Main Railway Station. Machine learning-based prediction models were developed by considering factors such as train type, density, and day of the week. Logistic Regression (LR), Support Vector Machines (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) algorithms were comparatively evaluated. After hyperparameter optimization, the XGBoost algorithm achieved the highest performance, with an accuracy of 64% and an F1-score of 68% on the test data. The results demonstrate the effectiveness of machine learning approaches in predicting train delays.