摘要
文章以云南为例,对同比月度CPI原始数据进行小波阈值去噪处理,对去噪后数据作BP神经网络预测,通过SARIMA模型修正预测残差,从而建立了小波分析的BP-SARIMA模型。实证结果表明：小波分析的BP-SARIMA模型相对于未经小波分析的BP-SARIMA模型及未经SARIMA残差修正的小波分析的BP模型有更高的预测精度。
This paper takes Yunan Province as example to de-noise wavelet threshold of monthly CPI raw data and predict the de-noised data by BP neural network, and then the paper uses the SARIMA model to correct predicted residuals and establish wavelet analytic BP-SARIMA model. The empirical results show that prediction accuracy of the wavelet analytic BP-SARIMA model is higher than that of the BP-SARIMA model without being processed by wavelet analysis and the BP model without residual correction by the SARIMA.
作者
彭乃驰
党婷
Peng Naichi;Dang Ting(Department of Information Science and Technology,Tourism and Culture College of Yunnan University,Lijiang Yunnan 674199,China)
出处
《统计与决策》
CSSCI
北大核心
2018年第16期22-25,共4页
Statistics & Decision
基金
云南省教育厅科学研究基金项目(2015Y507)
