摘要
研究了人工神经网络在经济预测中的应用问题,探讨了人工神经网络的时间序列预测方法。该方法采用多层前馈神经网络及BP算法,其仿真实现是以MATLAB下神经网络工具箱作为开发工具。文中提出了一种基于BP网络时序预测通用方法,并通过实例验证了该方法的预测精度明显高于灰色系统预测方法。为了消除单一神经网络预测模型的系统偏差,探讨了组合神经网络时序预测方法;并用实例验证了组合神经网络比单一神经网络的预测精度高。
This dissertation studies applied-problems of artificial neural network （ANN） in economic prediction. In this thesis, ANN approaches for time series prediction is discussed. In the method above, muhilayered feedforward neural network and BP training algorithm are adopted and neural network toolbox based on the software MATLAB is used for emulational realization.
In this paper, a standardized method of neural network model based on BP network for time series prediction is proposed. Next the fact is tested with the prediction example of time series that prediction precision of ANN method for time series is much higher than that of grey system method. In order to remove the systematic deviation of the single neural network model, the way of combined neural network for time series prediction is discussed and the advantage of this method is proved with the experiment.
出处
《价值工程》
2007年第5期90-93,共4页
Value Engineering
