改进RBF神经网络在城市公交车速时间序列预测中的应用

Application of Based on Modified RBF Neural Network in City Bus Speed Time Series Forecasts

针对城市公交车速在时序上的复杂非线性特征及目前预测方法的不足,采用径向基函数(RBF)神经网络对城市公交车速时间序列进行预测。在Matlab R2007b环境下,建立RBF神经网络城市公交车速预测模型,并应用于兰州市103路公交车的车速预测.同时对网络的输入变量进行优化改进,设计网络参数,进行网络学习与训练的数值仿真试验.对比改进的RBF神经网络与标准的RBF及具有动量梯度算法的BP神经网络的预测结果。结果表明,该模型拟合精度和预测精度较高、计算速度较快。

To resolve the timing characteristics of the complex nonlinear and shortcomings of current forecasting methods for the speed of city buses, Radial Basis Function （RBF） neural network is applied to predict the city bus speed time series. The Matlab R2007b based RBF neural network prediction model of city bus speed time series is set up, and applied to predict the speed of Lanzhou City bus speed of 103. While the input variables of the network is optimized to design the improved network parameters, the numerical simulation experiments of the network＇ s learning and training are done to compare the prediction results of the improved RBFNN with those of standard RBFNN and BPNN with momentum gradient algorithm. The results show that the improved RBFNN is more accurate and precise in fitting, and gains faster computation.