基于ARIMA-BP神经网络组合预测的港口吞吐量预测  

Port Throughput Prediction Based on ARMI-BP Neural Network Combination Prediction

作　　者：曹莹 陈旭 张跃博 龚正 CAO Ying;CHEN Xu;ZHANG Yuebo;GONG Zheng(Hubei Transportation Planning&Design Institute Co.,Ltd.,Wuhan 430051,China)

机构地区：[1]湖北省交通规划设计院股份有限公司,湖北武汉430051

出　　处：《物流技术》2023年第12期84-91,共8页Logistics Technology

摘　　要：为提高港口吞吐量预测精准性,建立了反向神经网络(BP)和差分整合移动平均自回归模型(ARIMA)的组合预测模型。首先考虑能够体现地方社会经济发展的经济评价指标,通过反向神经网络和差分整合移动平均自回归预测,分别得到港口吞吐量的预测结果;再运用拟合优度方法赋权组合,得到组合预测结果。以天门港为案例,组合预测模型的误差为0.072%,预测精度较高,未来可应用于短期水运工程预测。In order to improve the accuracy of port throughput prediction,we established a prediction model that combined back propagation neural network(BP)and differential autoregressive integrated moving average(ARIMA).Firstly,considering the economic evaluation indicators capable of reflecting local social and economic development,we obtained the port throughput prediction result respectively using BP and ARIMA.Then,we used the goodness of fit method to get the combination prediction result.Finally,taking Tianmen Port as an example,we found that the error of the combination prediction model was 0.072%,proving the high prediction accuracy of the combination prediction model.

关 键 词：港口吞吐量 组合预测 反向神经网络 差分整合移动平均自回归 天门港 

分 类 号：U652.14[交通运输工程—港口、海岸及近海工程]