Simpson改进的灰色神经网络在汽车保有量中的预测  

Prediction of China’s Car Ownership by Grey Neural Network with Simpson Formula

作　　者：吴文青[1] 夏杰 WU Wenqing;XIA Jie(School of Science,Southwest University of Science and Technology,Mianyang 621010,Sichuan,P.R.China;School of Mathematical Sciences UESTC,University of Electronic Science and Technology of China,Chengdu 611731,Sichuan,P.R.China)

机构地区：[1]西南科技大学理学院,四川绵阳621010 [2]电子科技大学数学科学学院,四川成都611731

出　　处：《重庆交通大学学报（自然科学版）》2019年第9期101-108,共8页Journal of Chongqing Jiaotong University(Natural Science)

基　　金：教育部人文社科青年基金项目(19YJCZH119);西南科技大学博士研究基金项目(15zx7141)

摘　　要：针对汽车保有量数据具有非线性和随机性的特点,建立基于Simpson公式的灰色神经网络模型对汽车保有量进行预测研究;利用Simpson公式对经典GM(1,1)灰色系统的背景值进行改进以提高模型的预测精度;通过相关性分析,确定国民总收入、人均国内生产总值、总人口、固定资产投资、进出口总额、钢材产量、社会消费品零售总额7个因素为汽车保有量的影响因素,并将7个影响因素作为BP神经网络的输入建立BP神经网路模型;根据灰色系统和BP神经网络预测误差大小确定组合模型的权重,构建灰色神经网络组合模型;对比分析经典GM(1,1)、Simpson公式的GM(1,1)、BP神经网络、灰色神经网络、Simpson公式的灰色神经网络模型的计算结果。研究表明:基于Simpson公式的灰色神经网络预测精度最高,其相对误差均在3%以内,相对误差的方差为3.2780,小于灰色神经网络模型和单一预测模型。According to the features of non-linearity and randomness of car ownership data,a grey neural network model with Simpson formula was established to predict China s car ownership.First,the formula was employed to construct the background value of classic GM(1,1)for improving prediction accuracy;second,such seven factors of car ownership as total gross national income,per capita gross domestic product,total population,investment in fixed assets,total import and export,steel output and total retail sales of consumer goods were viewed by correlation analysis and then the input of neural network was used for building BP neural network model.To determine the weight of the combined model with the prediction error of grey system and BP neural network,the grey neural network model was constructed.With the comparison among the classic grey GM(1,1),GM(1,1)of Simpson formula,BP neural network,grey neural network,Simpson formula grey neural network model,the results show that grey neural network with Simpson formula has the highest prediction accuracy where the relative error is as low as 3%and the mean square deviation is 3.2780%,which is smaller than grey neural network model and single prediction model.

关 键 词：车辆工程 汽车保有量 背景值 Simpson公式的GM(1 1)模型 组合预测模型 预测精度 

分 类 号：U491.4[交通运输工程—交通运输规划与管理]