Primary delays are the driving force behind delay propagation, and predicting the number of affected trains (NAT) and the total time of affected trains (TTAT) due to primary delay (PD) can provide reliable decision support for real‐time train dispatching. In this paper, based on real operation data from 2015 to 2016 at several stations along the Wuhan–Guangzhou high‐speed railway, NAT and TTAT influencing factors were determined after analyzing the PD propagation mechanism. The eXtreme Gradient BOOSTing (XGBOOST) algorithm was used to establish a NAT predictive model, and several machine learning methods were compared. The importance of different delayinfluencing factors was investigated. Then, the TTAT predictive model (using support vector regression (SVR) algorithms) was established based on the NAT predictive model. Results indicated that the XGBOOST algorithm performed well with the NAT predictive model, and SVR was the optimal model for TTAT prediction under the verification index (i.e., the ratio of the difference between the actual and predicted value was less than 1/2/3/4/5 min). Real operational data in 2018 were used to test the applicability of the NAT and TTAT models over time, and findings suggest that these models exhibit sound applicability over time based on XGBOOST and SVR, respectively.

中文翻译：

---

使用高速列车运行记录的主要延误​​影响的预测模型

主要延迟是延迟传播的驱动力，并且预测由于主要延迟（PD）导致的受影响列车的数量（NAT）和受影响列车的总时间（TTAT）可以为实时列车调度提供可靠的决策支持。本文基于武广高铁沿线几个站点2015年至2016年的实际运行数据，通过分析局部放电的传播机理，确定了NAT和TTAT的影响因素。使用eXtreme梯度提升（XGBOOST）算法建立NAT预测模型，并比较了几种机器学习方法。研究了不同延迟影响因素的重要性。然后，基于NAT预测模型建立TTAT预测模型（使用支持向量回归（SVR）算法）。结果表明，XGBOOST算法在NAT预测模型上表现良好，而SVR是在验证指标下TTAT预测的最佳模型（即，实际值与预测值之差的比率小于1/2/3 / 4/5分钟）。2018年的实际运营数据用于测试NAT和TTAT模型随时间推移的适用性，研究结果表明，这些模型分别基于XGBOOST和SVR表现出良好的适用性。