将小波分析和支持向量回归(SVR)模型引入消费者物价指数CPI的时间序列分析中,利用小波降噪对原始时间序列进行小波变换,充分提取和分离金融时间序列的各种隐周期和非线性,把小波分解序列的特性和分解数据随尺度倍增而倍减的规律充分用于SVR支持向量回归模型的建模。将该方法应用于中国宏观经济指标CPI的分析与预测,可以有效预测CPI的变动方向,并显著提高CPI的预测精度。
This paper applies wavelet and support vector regression (SVR) to study various time series properties of Chinese consumer price index (CPI).Our model can forecast Chinese CPI more accurately than popular single-period models in the literature based on root mean square error(RMSE)mean absolute error(MAE)and direction accuracy(DA).