论文部分内容阅读
本文提出基于改进自组织方法的GMDH(Group Method of Data Handling)型神经网络并将它应用于混沌预测。一般的GMDH型神经网络的自组织功能是通过给定一个准则阈值来确定或直接给定数值来实现,但GMDH型神经网络的自组织准则的阈值难以合适确定,由此提出了一种简单的自组织方法来实现真正意义上的自组织功能。这种用改进了的自组织方法所构成的GMDH型神经网络可以应用于混沌时间序列预测。通过仿真实验,证明其预测效果明显比基本的GMDH型神经网络好,即改进GMDH型神经网络优于基本的GMDH型神经网络。
In this paper, GMDH (Group Method of Data Handling) neural network based on improved self-organizing method is proposed and applied to chaos prediction. The general self-organizing function of GMDH neural network is achieved by giving a criterion threshold or by directly giving a given value. However, the threshold of self-organizing criterion of GMDH neural network is difficult to be determined properly. Therefore, a simple Self-organizing approach to achieve the true self-organizing function. The GMDH neural network constructed by the improved self-organizing method can be applied to chaotic time series prediction. The simulation results show that the prediction effect is better than the basic GMDH neural network, that is, the improved GMDH neural network is better than the basic GMDH neural network.