基于VLRBP神经网络的汇率预测

为了提高汇率预测的准确性,分别使用VLRBP神经网络模型和GRNN模型及ARIMA模型对欧元汇率时间序列进行建模和预测,通过实证分析发现基于VLRBP的神经网络对于含有大量非线性成分的欧元汇率时间序列的预测比较准确.在分析了最速下降BP学习算法的缺点后,提出利用VLRBP学习算法来解决神经网络振荡和收敛速度过慢的缺陷,并取得较好的效果.同时,为了提高VLRBP网络的泛化性能,提出在训练VLRBP神经网络时应用浴盆曲线方法选取隐层神经元个数和滑动窗口尺寸,试验结果表明该方法适合神经网络模型.

Foreign exchange rates forecasting based on VLRBP artificial neural networks

To improve the performance of the foreign exchange rates forecasting,a neural networks model and an ARIMA model are built,and then the euro foreign exchange rate is studied by using the two models.The result shown that the BP neural networks produced better performance than the other methods in forecasting the euro foreign exchange rate which had a great amount nonlinear components.Based on the analysis of the deficiency of the SDBP algorithm,the VLRBP algorithm is presented to solve the oscillation and slo...