基于神经网络的交通参数预测方法

为能够迅速准确地采取相应措施处理交通拥堵问题,改善行车安全,进而提高路网效率,研究了基于神经网络的交通参数预测方法,预测了交通流量、速度和占有率.在分析常用BP(Back Propagation)神经网络算法的基础上,研究误差平方和最小化的L-M(Levenberg-Marquart)算法.相对于常规预测方法,基于神经网络的交通参数的预测方法对于随机的参数变化具有更好的适应性,能及时跟随交通参数的变化,所以精确度更高,适应性更好.仿真结果显示,L-M算法的训练速度相比于常规BP算法要快几十倍,预测交通流量、速度和占有率等参数的效果优于常用的指数平滑算法,因此基于神经网络的交通参数预测方法可以应用于交通领域.

Algorithm of Traffic Parameters Prediction Based on Neural Network

To deal with the traffic jam,promote the safety level of the traveling,and improves the efficiency of limited road network.Based on the neural network.We can use the algorithm of traffic parameters prediction to predict the traffic flow, speed and proportion.On the base of analyzing the BP(Back Propagation) neural network algorithm,we study on the L-M(Levenberg-Marquart) algorithm,which error square sum is minimum.Compared with the general predicted algorithms,the algorithm of traffic parameters prediction based on neural network is superior on treating with stochastic parameters.It can follow the traffic parameters.So it is more exact and adaptive.The simulation results show that the training speed of the L-M algorithm is fast than that of the BP algorithm.That of predicting the traffic flow,speed and proportion is better than the effect of exponential smoothing algorithm.Consequently,the algorithm of traffic parameters prediction based on Neural Network can be applied in traffic field.