基于改进WNN的短时交通流量预测的研究  

Research on Short-term Traffic Flow Prediction Based on Improved WNN

作　　者：沈子华 冯益春 朱媛媛[1] 王笑梅[1] 郭超磊 SHEN Zihua;FENG Yichun;ZHU Yuanyuan;WANG Xiaomei;GUO Chaolei(College of Information,Mechanical and Electrical Engineering,Shanghai Normal University,Shanghai 200030)

机构地区：[1]上海师范大学信息与机电学院,上海200030

出　　处：《计算机与数字工程》2020年第7期1617-1622,共6页Computer & Digital Engineering

摘　　要：论文提出了一种结合自适应人工鱼群算法及粒子群优化算法的混合寻优算法,替换了小波神经网络在求解最优参数时所使用的梯度下降法,从而可在一定程度上优化小波神经网络。在此基础上,采用改进的小波神经网络对短时交通流量数据进行建模与预测。仿真实验结果表明,在进行短时交通流量预测时,改进后的小波神经网络模型相比其他一些主流的优化预测模型,无论在收敛速度、稳定性还是在预测精度方面均有一定的提高。In this paper,a hybrid optimization algorithm combining adaptive artificial fish swarm algorithm and particle swarm optimization algorithm is proposed to optimize the prediction model of wavelet neural network by replacing the gradient descent method used by wavelet neural network to solve the parameter optimal solution.On this basis,the improved wavelet neural network is used to model and predict the short-term traffic flow data.The simulation results show that,in short time traffic flow forecasting,compared with some other mainstream optimization prediction models,the improved wavelet neural network model proposed in this paper has a certain improvement in terms of convergence rate,stability and prediction accuracy.

关 键 词：小波神经网络 梯度下降 自适应人工鱼群算法 粒子群算法 短时交通流量 

分 类 号：U491[交通运输工程—交通运输规划与管理] TP183[交通运输工程—道路与铁道工程]