Abstract: The main purpose of this paper is to investigate the application of the neuron network (NN) for the daily exchange rate forecasting. Generalized Cross Validation (GCV) is introduced to determine the number of nodes of the hidden layer, several well known time series forecasting methods are also compared with the NN method in this paper.

Key words: artificial neuron network, generalized cross validation, exchange rate forecasting

摘要： 讨论了人工神经网络在金融汇率预报中的应用。其中介绍了广义交互验证(Generalized Cross Validation)法如何应用于确定神经网络中隐层的个数，并用实例说明了该方法甚至对复杂的非线性函数也可以得到很好的逼近。详细地介绍了运用人工神经网络作两周向前汇率预报的计算步骤。其平均相对误差(APE)为10*E-3的数量级，而国际上通用的状态空间模型及Box-Jenkins的ARIMA模型的预报误差都在10*E-2的数量级。

关键词: 人工神经网络, 广义交互验证法, 汇率预报