基于小波神经网络和ARMA模型的CPI波动区间预测  

Based on wavelet neural network and ARMA model of CPI fluctuation range forecast

作　　者：张泼泼 李星野[1] 

机构地区：[1]上海理工大学管理学院,上海200093

出　　处：《信息技术》2016年第2期131-135,共5页Information Technology

摘　　要：目前对于消费者价格指数(CPI)的预测研究基本集中于点预测。为预测本期较上期的CPI数据波动区间,提出一种基于小波神经网络和ARMA组合模型预测的方法。该模型首先利用小波神经网络对CPI数据进行拟合测试,对测试序列实际输出和期望输出的残差序列{et}进行ARMA建模预测,然后基于方差最小原则得到预测残差序列{e!t}95%的置信区间。通过实验表明预测残差序列95%置信区间可以很好的反应未来CPI数据的波动情况,具有较高的参考价值。So far,a lot of research is focused on the prediction of the consumer price index（ CPI）. This paper presents a combination model based on wavelet neural network and ARMA model for forecasting the range of CPI volatility. Firstly,the model is used to fit the CPI sequence test and the test sequence of the actual output and desired output residual sequence of ARMA model building simulation. Then based on the principle of minimum variance the residual error sequence prediction interval is 95%. The experiment result shows that confidence interval can forecast CPI data volatility under fixed variance 95%confidence interval,which has high reference value.

关 键 词：CPI 小波神经网络 ARMA 置信区间 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] O212.1[自动化与计算机技术—控制科学与工程]