Cause-based Distribution Models of Affected Trains Account for Primary Delay in High-Speed Rail

To measure the severity of high-speed train delays, the distributions models of the affected trains, owing to different cause-based primary delays, were established using statistical modeling. First, the train operation records during 2014 and 2015 obtained from China Railway Guangzhou Group Co. Ltd. were considered to train the models, and the goodness of fit of the five candidate models was compared. Subsequently, the inverse regression model was used to model the distributions of the cause-based affected trains, and the parameters of the models were estimated using the R-program. Finally, the chi-square test was performed, and subsequently, the Kolmogorov-Smirnov double sample tests with respect to the models were employed based on the testing data of 2016. The test results indicate that all the cause-based models can pass the accuracy test under the significance level of 0.05, and the established models are identically distributed according to the testing data. The tests also indicate that the matching degree between the prediction results and the observations owing to the delay causes can be higher than 97%.