Prediction of high-speed railway train delay evolution based on machine learning
Zhang Pu1,2, Meng Lingyun2, Li Baoxu3
1. State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing 100044; 2. School of Traffic and Transportation, Beijing Jiaotong University, Beijing 100044; 3. China Railway Shenyang Bureau Group Co. Ltd, Shenyang 110001
Abstract：Delay evolution is the whole process of delay generation, propagation, continuation, deterioration and recovery during train operation. Due to its characteristics such as high speed, high density and high frequency, the delay evolution of high-speed railway train is very complex, and it is difficult to carry out effective research through traditional methods. With the help of machine learning theory and method, this paper carries out data preprocessing methods such as feature standardization, sample expansion and sampling on Beijing-Shanghai high-speed railway train operation performance data and disperse the delay time into different ranges. A single train delay prediction model and a all-train delay prediction model based on support vector machine (SVM) are established and programmed. The validity of the model is evaluated by the constructed evaluation index. The results show that the model can predict the range of the delays ideally. This paper is an attempt of machine learning theory to study high-speed railway train delay evolution. Case studies prove that the new method is feasible in the prediction of high-speed railway train delays.
2019, Vol. 20 
