融合麻雀搜索与长短时记忆的疆煤铁路外运量预测  

Prediction of Outbound Transportation Volume of Xinjiang Coal Railway by Integrating Sparrow Search with Long Short-Term Memory

作　　者：李海军[1,2] 张晓洋 高如虎 魏德华 陈晓明 LI Haijun;ZHANG Xiaoyang;GAO Ruhu;WEI Dehua;CHEN Xiaoming(School of Transportation,Lanzhou Jiaotong University,Lanzhou 730070,China;Key Laboratory of Railway Industry on Plateau Railway Transportation Intelligent Management and Control,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学交通运输学院,兰州730070 [2]兰州交通大学高原铁路运输智慧管控铁路行业重点实验室,兰州730070

出　　处：《交通运输系统工程与信息》2024年第5期14-23,共10页Journal of Transportation Systems Engineering and Information Technology

基　　金：国家自然科学基金(72361020);中国国家铁路集团有限公司科技研究开发计划项目(K2023X019)。

摘　　要：为准确预测新疆煤炭铁路外运量,本文提出一种融合麻雀搜索算法与长短时记忆网络的(SSA-LSTM)预测模型,模型引入麻雀搜索算法对长短时记忆网络(Long Short-Term Memory,LSTM)模型的超参数进行优化,以提高模型预测性能。以2015—2022年新疆煤炭铁路外运量数据为基础,综合考虑经济、运输等多种因素,对各影响因素的灰色关联度进行计算,验证所选因素与预测指标具有较强的关联度。对影响因素中的GDP数据进行消费者价格指数(Consumer Price Index, CPI)处理,并将处理后的数据输入模型进行预测,最后应用该模型对未来新疆煤炭初、近、远期铁路外运量进行预测。研究结果表明,SSA-LSTM模型的预测效果显著优于BP(Back Propagation)神经网络和传统LSTM模型,平均绝对百分比误差(MAPE)为0.88%,均方根误差(RMSE)为49.9。同时,与未经CPI处理的预测相比,经过CPI处理后预测误差更小,MAPE和RMSE分别降低了75.8%和56.2%。本文为新疆煤炭铁路外运量预测提供了一种有效方法,为疆煤外运通道设计提供了重要数据支撑。To enhance the precision of predicting the Xinjiang's coal railway outbound volume transportation,a prediction model integrating the sparrow search algorithm and the long and short-term memory network(SSA-LSTM)is proposed.The model introduces the sparrow search algorithm to optimize the hyper-parameters of the LSTM model in order to improve the model prediction performance.Based on the data of Xinjiang coal rail outbound transportation volume from 2015 to 2022,the gray correlation analysis is employed to comprehensively evaluate the impact of factors,including economic and transportation aspects,ensuring that the selected factors exhibit a strong correlation with the prediction targets.Among the influencing factors,the GDP data is adjusted for Consumer Price Index(CPI)effects,and the refined data are then fed into the model for prediction.Finally,the model is applied to predict the Xinjiang's coal rail outbound transportation volume across short,medium,and long time horizons.The results demonstrate that the SSA-LSTM model outperforms both the BP neural network and the conventional LSTM model,achieving a Mean Absolute Percentage Error(MAPE)of 0.88%and a Root Mean Square Error(RMSE)of 49.9.Furthermore,incorporating CPI processing into the prediction process significantly reduces the prediction error,with MAPE and RMSE decreasing by 75.8%and 56.2%,respectively,compared to non-CPI-processed predictions.This study provides an effective approach for predicting Xinjiang's coal rail outbound transportation volume,offering important data insights that inform the strategic design of coal transportation routes out of Xinjiang.

关 键 词：铁路运输 疆煤铁路外运量 SSA-LSTM模型 灰色关联分析 预测 

分 类 号：U294.13[交通运输工程—交通运输规划与管理]