Due to various reasons, variabilities in freight train travel time may be significant, yielding considerable challenges to forecasting the estimated time of arrival (ETA). Based on historical railway ETA data, this study first analyzes the shared route, then converts the historical data information of a train into multiple time data points, and finally, builds a tree-based model selection framework using the random forest algorithm (RF) and a feature weighted K-nearest neighbor algorithm (FWKNN) to create a phased prediction model. In terms of time, we study the use of different algorithms to predict the ETA of freight trains at various locations on freight train routes. In this study, the proposed method was tested on the dataset of the 2021 The Institute for Operations Research and the Management Sciences (INFORMS) Railway Applications Section (RAS) problem solving competition and won 2nd place.
Forecasting estimated times of arrival of US freight trains
Transportation Planning and Technology ; 45 , 5 ; 427-448
2022-07-04
22 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Unbekannt
British Library Online Contents | 2000
|Engineering Index Backfile | 1907
Engineering Index Backfile | 1928