基于遗传算法优化小波神经网络的短时交通流量预测

SHORT-TERM TRAFFIC FLOW FORECASTING BASED ON WAVELET NEURAL NETWORK OPTIMISED BY GENETIC ALGORITHM

为了提高小波神经网络对具有时变性、非线性和复杂性等特点的短时交通流量预测的准确性,提出一种基于遗传算法优化小波神经网络的短时交通流量预测模型。利用遗传算法隐含并行性、自适应随机搜索及全局寻优的特性,优化小波神经网络的权值和阈值,克服了小波神经网络易陷入局部最优、得不到最优参数的缺陷。仿真结果表明,该方法对短时交通流量具有较好的非线性拟合能力和更高的预测精度,并具有良好的应用价值。

In order to increase the accuracy of using wavelet neural network（ WNN） to forecast short-term traffic flow,which is time-variant,nonlinear and complex,a prediction method for short-term traffic flow of optimized WNN based on genetic algorithm（ GA） was proposed. The weights and thresholds of wavelet neural were optimized by using the characteristics of the genetic algorithm,such as implicit parallelism,adaptive random search and global optimization.The defect of WNN,including easily falling into the local optimum and optimal parameters being unavailable,can be overcome. The simulation results show that this method has good nonlinear fitting ability and higher prediction accuracy for short-term traffic flow,and has significant application value.