基于预报-校正法的汇率预测模型

神经网络已成为金融时间序列预测的一个有力工具，但有些设计因素对神经网络的预测效果有很大的影响，这些因素包括输入变量选择、网络的结构和训练数据量。提出了基于预报一校正方法的神经网络预测模型，并对不同大小的训练集的影响进行了实验研究。结果发现大的训练集有更好的预测效果，且该方法的预测精度要普遍高于单一神经网络所能达到的效果。

Exchange Rate Forecast Model Based on Prediction-Revise Method

Neural network is a powerful tool for forecasting financial time series, but several design factors significantly impact the result accuracy. These factors include selection of input variables, architecture of the network, and quantity of training data. In this paper, a neural network forecast model based on the prediction-revise method is presented and the effects on different sizes of training sample sets on forecasting exchange rates are examined. The results show that the forecast effect on big training sample set is better than that on small set and the predication accuracy of the proposed approach is generally better than that of individual neural networks.