Forecasting and Analysis of Train Delays and Impact of Weather Data using Machine Learning

The Railway transit is one of the dominant means of transport all over the world. But people who use the railway transport system are usually affected by the train being delayed from it's usual schedule. This study focuses on forecasting the train delay in India from the available historical data. India has the fourth largest railway network in the world, conveying in excess of eight billion travelers each year. Nonetheless, the passengers usually experience inconvenience because almost all the time, the trains do not run as per the scheduled time. The delay stems from various factors like severe weather conditions, seasonal requirements, railway policies, technical issues, delays accumulated from preceding trains, etc and this imposes considerable costs on railways as well as the travelers. Due to this reason, people are forced to look for an alternate option to travel. Therefore, forecasting delays of trains play an important role in the railway transit. It could help attract more people as they could plan their journey accordingly and it could also help the authority to minimize future delays. An effective supervised machine learning regression algorithm could be useful for estimating train delays and obtaining accurate results. Our main contribution in this paper is the study of the impact of weather on the train running status. As we progress, the various factors of weather like temperature, rainfall, wind, etc are incorporated and analyzed in detail. So in this paper, the main focus is identifying a regression algorithm that can accurately predict train delays by considering the weather conditions as well. We have used 6 different regression models (ridge, lasso, elastic-net, SVR, Gradient Boosting, and XGB Regression) and we have also tried to calibrate the hyperparameters of each regression algorithms to get the most accurate results.