基于组合神经网络模型的函数逼近方法

FUNCTION APPROXIMATION BASED ON COMBINED NEURAL NETWORK MODEL

在通用函数逼近定理基础上,介绍了一种将反向传播神经网络和径向基神经网络模型相接合的组合神经网络模型,并将该模型应用于上海证券指数的预测。仿真实验结果表明,该模型很好地减小了预测值和实际值之间的误差,预测效果也优于普通的反向传播神经网络模型。

In this paper we introduce one Combined Neural Network Model which is based on overall function approximation theorem, The model combines back propagation neural network with radical basis neural network. In order to identifv the validity of this model, we make use of this model in forecasting Shanghai Stock Index. The experiment shows that our model can effectively reduce the bias of forecasting value against the real value and its forecasting results are also superior to common back propagation neural network models.