基于MIV-GA-BP神经网络的我国棉价预测研究  

Forecast of CN Cotton B Based on MIV-GA-BP Neural Network

作　　者：吴叶 刘婷婷[1] 方少勇[1] WU Ye;LIU Tingting;FANG Shaoyong(Beijing Forestry University, Beijing, 100083)

机构地区：[1]北京林业大学,北京100083

出　　处：《棉纺织技术》2018年第7期77-80,共4页Cotton Textile Technology

摘　　要：探究了MIV-GA-BP神经网络模型对我国棉花价格预测的情况。以国家棉花价格指数B作为棉价反映指标,选取了棉花产量、进口量、消费量等13个影响棉花价格的因素,采用平均影响值(MIV)、遗传算法(GA)与BP神经网络相结合的方法,按照15%的淘汰率进行筛选,得出我国棉价波动的主要影响因素,并在此基础上构建了MIV-GA-BP神经网络模型。以2015年1月—2017年12月的3128B棉花月度平均价格为样本数据,进行MIV-GA-BP模型拟合精度评估和预测精度评估。认为:基于MIV-GA-BP神经网络模型拟合精度良好,预测精度较高,训练样本可反映99%的样本特征。The situation of CN Cotton Bby using MIV- GA-BP neural network model was discussed. The na- tional CN Cotton B was used as reflection index of cotton prices. 13 factors affecting cotton prices such as cotton output,import output and consumption output were selected. The combination method of mean impact value （MIV） ,genetic algorithm （GA） and BP neural network was screened according to the elimination rate of 15%. Main influencing factors of CN Cotton B were obtained. MIV-GA-BP neural network model was constructed on this basis. The monthly average data of 3128B cotton price from January 2015 to December 2017 were taken as sample data, fitting precision and forecasting accuracy of MIV-GA-BP neural network model were evaluated. It is considered that fitting accuracy of MIV-GA-BP neural network model is better and prediction accuracy of the model is higher. The training sample can reflect 99G sample characteristics.

关 键 词：国家棉花价格指数 BP神经网络 平均影响值 变量筛选 价格预测 

分 类 号：TS101.8[轻工技术与工程—纺织工程]