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 -
The main purpose of this paper is to investigate the statistical analysis of J\|effect in trade balance of Mexico in 1989\|1995.Sparse coefficients modeling has been successfully introduced for the model construction of J\|effect.Based on the model,the recovery period index is easy to define and the policy experiment also may be carried out by orthogonal experiment design.The results show that the develuation of Pesos and the short interest rate are over adjusted but the economic and financial policy during that period were basically correct.\;Finally,J\|effects between Mexico and Japan are compared,it seems that the recovery period of developing countries will be longer than the industry countries.
