基于小波分析与神经网络的交通流短时预测方法

The Forecasting Approach for Short-term Traffic Flow Based on Wavelet Analysis and Neural Network

提出基于小波分析与神经网络的交通流短时预测方法,把多维输入进行小波分解降维,预测由多个子网络独立完成,有效解决了多维神经网络的映射学习容易产生"维数灾"的问题.示例结果表明,该方法比典型的神经网络预测准确度高、误差小.

The method based on wavelet analysis and neural network for short-term traffic flow forecasting is presented. The multidimensional inputs are decompounded by wavelet analysis and the forecasting is implemented by several sub neural networks independently. It resolves the dimension-disaster problem effectively in multidimensional neural network mapping. The demonstration results show that the method can evidently decrease prediction error and improve forecasting veracity compared with typical neural network.