误差修正Stacking算法在港口吞吐量预测中的应用  

Application of Stacking Algorithm for Error Correction in Port Throughput Prediction

作　　者：李朝辉 殷铭 王晓倩 张琳 LI Zhao-hui;YIN Ming;WANG Xiao-qian;ZHANG Lin(School of Maritime Economics and Management,Dalian Maritime University,Dalian Liaoning 116026,China)

机构地区：[1]大连海事大学航运经济与管理学院,辽宁大连116026

出　　处：《计算机仿真》2022年第2期512-517,共6页Computer Simulation

基　　金：教学部产学合作协同育人项目(201802028013,201902250011);辽宁省社会科学规划基金项目(L18CTQ004);中国博士后科学基金资助项目(2015M571292);中央高校基本科研业务费专项资金资助项目(3132020244)。

摘　　要：吞吐量的准确预测对于港口的发展具有重要意义,传统的单一模型在预测的精度或稳定性上都存在局限性。为提高吞吐量预测精度,引入Stacking算法的思想建立组合模型,并在此基础上提出了利用PCA-BP模型修正Stacking算法误差的误差修正模型;同时以1998-2017年深圳港集装箱吞吐量为实证研究对象,进行了基于BP神经网络和NARX神经网络单一预测模型、误差修正Stacking算法的预测模型及其它组合模型的预测对比分析,结果显示误差修正Stacking算法模型预测精准度最好,验证了模型的可行性,为港口的吞吐量预测提供参考。The accurate prediction of throughput is of great significance to the development of the port. The traditional single model has limitations in prediction accuracy or stability. In order to improve the accuracy of throughput prediction, the idea of the Stacking algorithm was introduced to establish a combined model, and on this basis, the PCA-BP model was used to correct the error of the Stacking algorithm. At the same time, the container throughput of Shenzhen Port from 1998 to 2017 was taken as an empirical research object. The prediction comparison analysis of the model shows that the error-correcting Stacking algorithm model has the best prediction accuracy, verifies the feasibility of the model, and provides a reference for the port’s throughput forecast.

关 键 词：神经网络 吞吐量预测 误差修正 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]