Prediction of high-speed train delay propagation based on causal text information

Abstract The delay-causing text data contain valuable information such as the specific reasons for delay, location and time of disturbance, which can provide an efficient support prediction train delays improve guidance control efficiency. Based on operation Wuhan–Guangzhou high-speed railway, relevant algorithms in natural language processing field are used to process data. It also integrates operating-environment so develop a cause-based delay propagation model. Word2vec model is first vectorize description after word segmentation. mean or term frequency-inverse document frequency-weighted then generate sentence vector based original vector. Afterward, features input into extreme gradient boosting (XGBoost) regression algorithm In this work, 4 feature methods 8 considered. results demonstrate that XGBoost has highest accuracy using test processed by continuous bag words models. Compared with only considers train-operating-environment features, show significantly improved multiple integrating feature.