粗正交小波网络及其在交通流预测中的应用

基于交通流预测的特点和输入向量的主成分分析方法,把粗集理论与正交小波网络相结合,给出了一种基于粗集的正交小波网络交通预测模型,并成功应用于城市交通流的实时预测.粗正交小波网络具有极强的鲁棒性,可以有效克服季节、天气等随机因素对交通流量预测性能的影响;主成分分析方法解决了正交小波网络多维输入时的维数灾难.实验结果表明,该模型的预测精度和收敛速度明显优于常规BP网络和小波框架神经网络,对交通流量等预测问题具有较高的应用价值.

Rough Orthogonal Wavelet Network and Its Applications to the Traffic Flow Forecast

Based upon the characteristics of traffic flow forecast and the principal component analysis (PCA), a new traffic flow forecasting model combined rough sets and orthogonal wavelet network together, called rough orthogonal wavelet network forecasting model, is put forward. The model has been successfully used to forecast urban traffic flow. Using the principal component analysis about input vectors, the model keeps away from the dimension avalanche of orthogonal wavelet networks. The experiment results show that the model is superior to the BP networks and wavelet frame networks in the aspects of flow forecasting precision and network convergence. The rough neural network forecasting model is robust to the uncertain factors for the traffic flow forecast. The model given in this paper is of academic and practical value in forecasting applications.