论文部分内容阅读
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 (GeneralizedCrossValidation)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 Jen kins的ARIMA模型的预报误差都在 10 E - 2的数量级
The application of artificial neural network in the forecast of financial exchange rate is discussed. It introduces how the Generalized CrossValidation method can be used to determine the number of hidden layers in a neural network. An example is given to show that this method can be well approximated even for complex nonlinear functions. The calculation steps of forecasting the exchange rate of two weeks ahead using artificial neural network are introduced in detail. The average relative error (APE) is on the order of 10E - 3. However, the prediction errors of the internationally accepted state space model and the Box Jenkins’ ARIMA model are all on the order of 10E - 2