Infrastructure – Aware Railway Delay Prediction Using Machine Learning and Civil Engineering Features

Abstract

Predicting railway delays is crucial because they can impact operations, lower passenger satisfaction, and pose safety risks. A machine learning system that forecasts train delays and identifies potential technical causes is presented in this study. Six thousand train records with sixteen features—train type, journey distance, schedule time, speed, bridge type, soil condition, season, flood risk, and rail vibration—were used to train the model. A variety of machine learning models, such as Support Vector Regressor, Random Forest, Gradient Boosting, Decision Tree, and Linear Regression, were examined. With a mean absolute error of 1.45 minutes and an R2 score of 0.95, the Gradient Boosting model produced the best results out of all of them, attaining an accuracy of almost 95%. A rule-based module of the system also describes potential engineering reasons for delays, including bridge speed limitations, soil settlement challenges, flood hazards, heat-induced rail expansion, and track vibration problems. To assist railway employees in making better decisions on traffic control and maintenance, the anticipated delays are divided into Low, Medium, and High categories. All things considered, this method improves railway reliability and allows for real-time railway monitoring by combining precise prediction with understandable explanations..