基于突发公共卫生事件影响下的铁路客流量恢复率预测研究

Prediction of Railway Passenger Flow Recovery Rate under the Influence of Public Health Emergencies

2020年以来新冠疫情随机性、区域性的爆发对铁路客流量的影响不容忽视。基于XGBoost算法建立城市间铁路客流量恢复率预测模型,提出以恢复率为预测目标,对疫情的严重程度和客流量的变化规律进行量化分析,为客流量预测和辅助决策提供参考,对于减少铁路客运收益损失具有重要价值。模型以所有二级及以上城市间铁路里程、城市间客流量、地理和铁路分布特征、受疫情影响程度等属性进行主成分分析,并进行维度压缩,获取新的输入变量。模型选取了2021年的数据随机采样,划分为训练集和测试集,采用5折交叉验证,并基于网格搜索法进行参数搜索得到最优模型参数并对测试集的恢复率进行了预测,与朴素贝叶斯和LightGBM算法进行对比,实验表明XGBoost算法的预测准确率较高。

Since 2020,the influence of random and regional COVID-19 outbreaks on railway passenger flow cannot be ignored.Based on the XGBoost algorithm,this paper built a prediction model for the recovery rate of inter-city railway passenger flow and employed the recovery rate as the prediction target to quantitatively analyze the pandemic severity and change laws of passenger flow.It provides references for passenger flow prediction and auxiliary decision-making and is of significance for reducing the revenue loss of railway passenger transport.The model conducted a principal component analysis based on attributes including the railway mileage among cities of level II and above,passenger flow among cities,geographical and railway distribution characteristics,and the influencing degree of the pandemic,and dimension compression was performed to obtain new input variables.Additionally,the model selected random data sampling in 2021,divided it into training sets and test sets,adopted five-fold cross-validation,and conducted parameter searches based on the grid search method to obtain the optimal model parameters and predict the recovery rate of the test set.Experiments show that the prediction accuracy of the XGBoost algorithm is higher than that of Naive Bayesian and LightGBM algorithms.