摘要
降雨给城市道路行程时间的计算和预测带来了许多不确定因素.以出租车GPS数据为研究对象,在考虑降雨数据的基础上,设计一个基于非最小路段的行程时间计算方法,建立基于LSTM(Long Short-Term Memory)循环神经网络的行程时间预测模型进行算法验证.最后,以北京市中关村西区出租车行驶的10 d的GPS数据进行方法验证.结果表明,加入降雨特征预测的结果比未加入降雨特征拥有更高的准确率.并与应用较为广泛的BP神经网络和SVM进行对比分析,发现在满足数据精度的前提下,本文应用的算法和预测模型有较高的训练速度和预测可靠性.
The precipitation brings many uncertainties to calculation and prediction of travel time in urban road.This paper used GPS data of taxi as the research,considered the precipitation data and then designed a travel time calculation method based on non-minimum section.Meanwhile,we had established a travel time prediction model based on the LSTM(Long Short-Term Memory)to verification the algorithm.Finally we used 10 days GPS data which is from taxi in zhongguancun west of Beijing to verify the method.The results show that the prediction results with rainfall characteristics are more accurate than those without.Compared with BP neural network and SVM witch are widely used,the algorithm and prediction model in our paper has higher training speed and prediction reliability under the premise of satisfying the accuracy.
作者
王志建
李达标
崔夏
WANG Zhi-jian;LI Da-biao;CUI Xia(Beijing Key Lab of Urban Intelligent Traffic Control Technology,North China University of Technology,Beijing 100144,China)
出处
《交通运输系统工程与信息》
EI
CSCD
北大核心
2020年第1期137-144,共8页
Journal of Transportation Systems Engineering and Information Technology
基金
国家自然科学基金(61503006)
北京市自然科学基金(8172018)
北京市教委基础科研计划项目(110052971921/023)~~
关键词
智能交通
旅行时间
LSTM神经网络
浮动车
降雨量
intelligent transportation
travel time
LSTM neural network
floating car
precipitation
