面向不确定物流需求的改进GM-BPNN组合预测方法  

An Improved GM-BPNN Combined Forecasting Method for Uncertain Logistics Demand

作　　者：黄建华 张迪 Huang Jianhua;Zhang Di(School of Economics and Management,Fuzhou University,Fuzhou 350108,China)

机构地区：[1]福州大学经济与管理学院,福州350108

出　　处：《统计与决策》2022年第16期26-29,共4页Statistics & Decision

基　　金：国家社会科学基金一般项目(20BGL003)。

摘　　要：在面对具有突变性、不稳定性以及非线性等特征的区域物流需求预测问题时,传统的时间序列、BPNN、GM-BPNN等预测方法在拟合物流需求曲线上存在缺陷,文章提出了改进GM-BPNN组合预测方法,利用ARIMA和遗传算法(GA)分别改进GM(1,1)和BPNN,根据有效度确定加权系数并构建线性组合模型,并以浙江、广东、江苏进行实例验证。结果表明,相比传统时间序列、BPNN、多元回归、GM-BPNN等预测方法,改进的GM-BPNN组合预测方法提高了物流需求预测的精确度。In the face of regional logistics demand forecasting with the characteristics of mutability, instability and nonlinearity, traditional time series, BPNN, GM-BPNN and other forecasting methods have defects in fitting the logistics demand curve.This paper proposes an improved GM-BPNN combined forecasting method, uses ARIMA and genetic algorithm(GA) to improve GM(1,1) and BPNN respectively, and determines the weighting coefficient according to the validity, with the linear combination model constructed. Finally, the model is verified by examples of Zhejiang Province, Guangdong Province and Jiangsu Province.The results show that compared with traditional time series, BPNN, multiple regression, GM-BPNN and other forecasting methods,the improved GM-BPNN combined forecasting method improves the accuracy of logistics demand forecasting.

关 键 词：区域物流需求预测 改进GM-BPNN ARIMA 遗传算法 组合预测 

分 类 号：F252.21[经济管理—国民经济]