摘要
利用支持向量机方法对汇率进行预测是金融市场研究领域一个重要的研究课题。结合小波变换与支持向量回归,提出一个三阶段时间序列预测模型。先以离散小波框架将汇率序列分解成不同尺度的多个子序列,揭示蕴含在预测变量内的信息,并对各个子序列进行时间序列分析,再以支持向量回归为工具,以这些子序列为预测变量建构支持向量回归模型,最后将各个序列的预测结果进行重构,得到预测结果。实证结果显示,该模型的预测效果较之BP神经网络与单纯的AR-SVM模型更优,证明基于小波分析与支持向量机相结合的预测模型可以为人民币兑美元汇率提供比较准确的预测。
In the field of financial market behavior research, forecasting exchange rate with SVM is an important problem. This paper brings forward a three - stage model combining with wavelet transform and support vector regression to predict exchange rate. By wavelet transform the exchange rate series are decomposed into several sub series to make the information hidden in predict variable known. The sub series are predicted with auto -regression models to seek optimal lag periods, which are more predictive and less correlated. And then constructs a SVR forecasting model with these sub series. The prediction result of the original time series is the reconstruction of the respective prediction. The empirical results show that the proposed model outperforms the AR - SVM model and BP neural network model. Experiments show the effectiveness of the predicting method. Therefore this prediction model can be effectively used in RMB/USD exchange rate series forecasting.
出处
《湘潭大学学报（哲学社会科学版）》
CSSCI
北大核心
2009年第5期82-87,共6页
Journal of Xiangtan University:Philosophy And Social Sciences
基金
国家社会科学基金重点资助项目(项目编号:07AJL005)
全国高校青年教师奖励基金资助项目(教2002[123])
关键词
人民币汇率预测
小波分析
滞后阶
支持向量机
forecasting RMB exchange rate
wavelet analysis
lag periods
support vector machine
