摘要
短期货运量预测研究是铁路运输企业编制日常工作计划的重要依据,准确的货运量预测结果对铁路货运组织工作具有积极意义。针对铁路短期货运量预测,建立基于双向长短时记忆网络(Bi-LSTM)的短期货运量预测模型,以某铁路局集团公司4122 d、136个月的货运发送量为实验数据分别进行各月和每日货运发送量的预测,其误差分别为5.30%和6.92%,并在同样的训练集、测试集数据分集上,设置相同的超参数,与RF,SVM,XGBoost和LSTM 4种模型的预测结果进行比较,验证Bi-LSTM网络在铁路短期货运量预测上的精确度和泛化能力较好。
Short-term freight volume prediction is an important basis for railway transportation enterprises to make daily work plans.Accurate freight volume prediction has positive significance for the organization of railway freight.In this paper,a model based on bi-directional long short-term memory(Bi-LSTM)was proposed to predict the short-term railway freight.According to the freight volume of a railway group within 4122 days or 136 months,this paper predicted the monthly and daily freight volume with errors of 5.30%and 6.92%respectively.Under the same training set and the data subset of the test set with the same hyperparameters,the Bi-LSTM network excels in the prediction accuracy of short-term railway freight volume and generalization ability in comparison with RF,SVM,XGBoost and LSTM.
作者
郭洪鹏
刘斌
肖尧
GUO Hongpeng;LIU Bin;XIAO Yao(School of Traffic and Transportation,Lanzhou Jiaotong University,Lanzhou 730070,Gansu,China)
出处
《铁道货运》
2022年第2期52-58,共7页
Railway Freight Transport
基金
国家自然科学基金项目(71761023)。
关键词
铁路运输
货运量
短期预测
双向长短时记忆网络
深度学习
Railway Transportation
Freight Volume
Short-term Forecast
Bi-LSTM Network
Deep Learning
