摘要
为了克服非线性和强噪声特征对交通流短时预测准确度的影响,应用交通流预测模型获得更为准确的交通流信息是智能交通建设的关键环节。文中构建了小波降噪的深度极限学习机对城市道路的交通流量进行预测,并与原极限学习机和小波BP神经网络模型的预测效果进行比较。将实验城市一年中电子警察采集到的各路口五分钟车流量作为训练集,构建了极限学习机、基于小波降噪的深度极限学习机和小波BP神经网络模型,分别对各路口高峰时段车流量进行预测,采用三类误差分析指标刻画三种模型的预测效果。实验结果表明,小波降噪的深度极限学习机预测误差评价值MAPE为0.234%,MRE为0.0029,RSE为0.6999,其值均小于原极限学习机和小波BP神经网络的误差指标,有较好的预测效果,从而说明小波降噪的深度极限学习机对短时交通流预测的合理性和可行性,为短时交通流的预测提供了一种新的解决思路。
In order to overcome the influence of nonlinear and strong noise characteristics on the accuracy of short-term traffic flow prediction,the application of traffic flow prediction model to obtain more accurate traffic flow information is the key to the construction of intelligent transportation system.In this paper,the deep extreme learning machine with wavelet de-noising is constructed to predict the traffic flow,and the prediction effect is compared with the original extreme learning machine and wavelet BP neural network model.Taking the five minute traffic flow of each intersection collected by the electronic police in one year as the training set,the extreme learning machine,the deep extreme learning machine with wavelet de-noising and the wavelet BP neural network model are constructed to predict the traffic flow of each intersection in peak hours,and three kinds of error analysis indexes are used to describe the prediction effect of the three models.The experiment shows that the MAPE,MRE and RSE of deep extreme learning machine wavelet de-noising are 0.234%,0.0029 and 0.6999 respectively,which are less than the error indexes of the original extreme learning machine and wavelet BP neural network,with better prediction effect.Therefore,the rationality and feasibility of deep extreme learning machine with wavelet de-noising for short-term traffic flow prediction are illustrated,which provides a new solution for short-term traffic flow prediction measurement.
作者
范馨月
FAN Xin-yue(Faculty of Mathematics,School of Mathematics and Statistics,Guizhou University,Guiyang 550025,China)
出处
《计算机技术与发展》
2021年第11期41-45,共5页
Computer Technology and Development
基金
贵州省科技计划项目(黔科合基础[2019]1122号)
贵州大学线上线下混合式课程建设项目(XJG202060)。
关键词
短时交通流预测
极限学习机
小波降噪
深度极限学习机
小波BP神经网络
short term traffic flow prediction
extreme learning machine
wavelet de-noising
deep extreme learning machine
wavelet BP neural network
作者简介
通讯作者:范馨月(1984-),女,博士,副教授,研究方向为大数据统计建模、计算数学。
