摘要
我国铁路货运量易受天气、节日和市场需求等众多因素的影响,使得铁路货运量具有周期性和波动性,预测难度高。本文综合考虑铁路货运量序列线性和非线性特征,建立SARIMA-PSO-ELM组合模型以提升预测的精度。首先使用SARIMA模型对我国铁路货运量序列进行预测,其次对SARIMA模型预测的残差建立PSO(粒子群优化)算法优化的ELM(极限学习机)预测模型,最后将两模型的预测值相加得到SARIMA-PSO-ELM组合模型的预测结果。组合模型预测的平均绝对误差(MAE)和平均绝对百分比误差(MAPE)分别是0.0129、0.35%,相较于SARIMA和PSO-ELM两种模型其预测精度更高。
China's railway freight volume is easily affected by many factors such as weather,festival and market demand,which makes the railway freight volume dificult to predict because of its periodicity and fuctuation.Considering the linear and nonlinear characteristics of railway freight volume series,the SARIMA-PSO-ELM combined model is established to improve the prediction accuracy.Firstly,using the SARIMA model to predict China's railway freight volume sequence.Secondly,ELM model is established by predicting the residuals of SARIMA model and optimizing them by the PSO(Particle Swarm Optimization)algorithm.Finally,the prediction results of the SARIMA model and the PSO-ELM model are combined to obtain the prediction results of the SARIMA-PSO-ELM combined model.The average absolute error(MAE)and average absolute percentage error(MAPE)of the combined model prediction are 0.0129 and 0.35%,respectively,which has higher prediction accuracy than the SARIMA and the PSO-ELM model.
作者
张仙
戴家佳
余奇迪
ZHANG Xian;DAI Jia-jia;YU Qi-di(Sdiool of Mathematics and Statistics,Guizhou University,Guiyang 550025,China)
出处
《数理统计与管理》
CSSCI
北大核心
2022年第3期394-401,共8页
Journal of Applied Statistics and Management
基金
贵州省数据驱动建模学习与优化创新团队(黔科合平台人才[2020]5016)。