结合粒子群算法的小波神经网络交通流预测

针对短时交通流量具有复杂性、非线性等特点,提出基于粒子群算法的小波神经网络交叉路口短时交通流量预测方法,利用粒子群算法优化小波神经网络的模型参数,通过定义可变的加速因子,使粒子群算法有利收敛于全局最优解.将粒子群算法的全局优化搜索能力和小波良好的时频局部性质相结合,克服神经网络易陷入局部极小和引起振荡效应现象的缺点.实验仿真结果说明,该算法可以有效提高预测精度,减少预测误差,并且很好的反映了交通流的特点.

Traffic Flow Forecasting Based on the Wavelet Neural Network with Particle Swarm Optimization Algorithm

The characteristic of short-term traffic flow is complexity and nonlinear.A new algorithm for crossroads short-term traffic flow forecasting based on the wavelet neural network with Particle Swarm Optimization algorithm is proposed,which uses Particle Swarm Optimization algorithm to optimize the initialized parameter of wavelet neural network.Through defining the variable accelerative factor,the search global optimization ability of Particle Swarm Optimization is improved.By combing Particle Swarm Optimization and the nice time-frequency local character of wavelet,the algorithm overcomes the phenomenon of network early falling into local minimal and oscillation effects.The results of simulation indicate that the algorithm can effectively improve the prediction precision,reduce the error,and depict the characteristic of traffic flow.