小波分析与非参数估计在汇率中的应用研究

Applied research of wavelet analysis and nonparametric estimation in exchange rate

首先,用小波Mallat算法对CHY/USD汇率一阶差分数据进行了分解和单支重构,得到了各层单支重构后的近似分量和细节分量;然后,基于局部线性非参数估计理论,对近似分量和细节分量分别建立了NARCH(1)模型;最后,对均值和波动率进行了10步预测.计算结果表明,非参数估计理论结合小波多分辨分析理论可以较好地应用于人民币汇率的预测,预测精度较高.

Firstly, decomposition and a single reconstruction of the first order difference data of CHY/USD exchange rate are made by the wavelet Mallat algorithm, and the approximate component and the detail component of a single reconstruction of the every level are obtained. Secondly, the nonparametric autoregression 1-D model are established to the approximate component and the detail component based on the local linear nonparametric estimation theory. Finally, a 10-step prediction is carried on to the average function and the volatility. The calculation results indicate that the nonparametric autoregression model combined wavelet can give better results for the CHY exchange rate prediction, and can be applied more effectively in the CHY exchange rate prediction.