基于随机森林的铁路冷藏运输需求预测  

Demand forecast of railway refrigerated transportation based on random forest

作　　者：夏伟怀[1] 刘嘉莉 冯芬玲[1] XIA Weihuai;LIU Jiali;FENG Fenling(School of Traffic and Transportation Engineering,Central South University,Changsha 410075,China)

机构地区：[1]中南大学交通运输工程学院,湖南长沙410075

出　　处：《铁道科学与工程学报》2022年第4期909-916,共8页Journal of Railway Science and Engineering

基　　金：科技部国家重点研发计划先进轨道交通重点专项资助项目(2018YFB1201402)。

摘　　要：为提升铁路冷藏运输效率和实现铁路冷藏运输资源的合理配置,推进铁路冷藏运输的快速发展,对铁路冷藏运输需求进行预测尤为重要。在分析铁路冷藏运输影响因素基础上,利用Spearman相关性分析进行特征筛选。结合Spearman相关性分析结果,构建基于随机森林的铁路冷藏运输需求预测模型,并与BP神经网络,AdaBoost,Bagging和未经特征筛选的随机森林预测模型的预测结果进行对比分析。研究结果表明:经Spearman相关性分析进行特征筛选后构建的随机森林回归预测模型的平均绝对误差和拟合优度值均优于其他模型,均方根误差值仅高于BP神经网络。随机森林回归预测模型的泛化能力较好,且特征筛选能够提高模型的预测精度。In order to improve the efficiency of railway refrigerated transportation, realize the reasonable allocation of railway refrigerated transportation resources, and promote the rapid development of railway refrigerated transportation, it is particularly important to predict the demand of railway refrigerated transportation.Based on the analysis of the influencing factors of railway refrigerated transportation, Spearman correlation analysis was used to screen the characteristics. Combined with Spearman correlation analysis results, a railway refrigerated transportation demand prediction model based on random forest was constructed. The prediction results were compared with those of BP neural network, AdaBoost, Bagging and random forest prediction models without feature selection. The results show that the average absolute error and goodness of fit of the random forest regression prediction model constructed by Spearman correlation analysis are better than other models, and the root mean square error is only higher than BP neural network. The developed random forest regression prediction model has good generalization ability, and feature screening can improve the prediction accuracy of the model.

关 键 词：铁路冷藏运输 需求预测 随机森林 特征筛选 

分 类 号：U295[交通运输工程—交通运输规划与管理]