基于GARCH模型与ANN技术组合的汇率预测

对汇率进行准确地预测是一项相当重要，也是十分困难的工作。到目前为止，人们已经提出了各种各样的方法和模型，但预测结果仍不能令人满意。近年来GARCH模型被广泛地用于对变动频率很高的金融时间序列建模，它能较好地抓住此类时间序列的动态特征。人工神经网络技术则是当前非常流行的一种替代传统的统计学模型，用来处理数据之间关系的技术，理论上它能以任意精度去逼近任意映射关系。将这二者结合起来对有关日汇率进行预测，获得了较好的预测表现。

Exchange Rate Forecasting Based on the Combination of GARCH Models and ANN Technologies

It is an important task, as well as a hard one, to give an accurate exchange rate forecasting. By now, all kinds of methods and models have been created, but those forecasting performances disappoint us. At present, GARCH type models have been employed to model these high frequency financial time series due to their ability to capture the dynamic characteristics. In addition, the artificial neural network (ANN) method has emerged as an alternative approach to statistical methodologies because it can theoretically approximate any mapping relationship to any desired degree of accuracy. The two methods are combined to forecast seven foreign exchange rates, and expected to have better performance.