基于GA优化IWNN的短时交通流量预测方法  

Short-time Traffic Flow Forecasting Method Based on Improved Wavelet Neural Network Optimized by GA

作　　者：吴凡[1] 孙建红[1] 葛鹤银 刘景夏[2] 

机构地区：[1]南京理工大学电子工程与光电技术学院,南京210094 [2]中国人民解放军理工大学通信工程学院,南京210007

出　　处：《实验室研究与探索》2016年第5期134-137,212,共5页Research and Exploration In Laboratory

基　　金：江苏省产学研联合创新基金(BY2014004)

摘　　要：由于交通流量的非线性、复杂性和不确定性,确定数学模型的预测方法难以满足交通管理控制中对预测精度和收敛速度的要求。为了对交通流进行准确、实时、高效的预测,提出将小波理论与神经网络相结合,并改进网络的训练过程从而构建改进型小波神经网络;同时运用遗传算法优化网络的初始权值,最终提高了预测精度,加快了收敛速度,避免陷入局部极小。通过仿真和分析,提出的方法具有较好的预测结果。Traffic flow prediction is a very important research area of intelligent transportation system,and has a very important academic value and practical significance to improve the traffic congestion problems. Traditional prediction methods which used determined mathematical model would not meet the needs of prediction accuracy and convergence speed during the traffic management control because of nonlinear,complexity and uncertainty of traffic flow. In order to forecast traffic flow accurately,real-timely and efficiently,a new algorithm is proposed by combining wavelet theory and neural network,and constructing an IWNN（ improved wavelet neural network） with improved network training methods.At the same time,the initial weights are optimized by GA（ genetic algorithm）. It can improve prediction accuracy,speed up the convergence speed and avoid entering local minima. The simulation results show that it can get better prediction results.

关 键 词：交通拥堵 短时交通流量预测 改进型小波神经网络 遗传算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]