摘要
为了更加精准实时地发布道路交通状态以及交通流趋势预警,基于动态规划和剪枝技术的PELT变点检测方法,针对卡口数据进行交通流状态检测与短期预测研究。首先,对于交通流近平稳状态的检测,结合时间序列的ADF检验构建了交通流近平稳状态检测方法。该方法能够快速准确地将断面交通流序列划分为多个可能接近平稳状态的候选间隔,接着对每个候选间隔进行ADF检验,在给定显著性水平下判别候选间隔是否平稳,进而检测出交通流序列中的近平稳状态。其次,对于交通流的短期预测,在PELT变点检测基础上将局部多项式回归与Sigma原则结合,得到短期断面交通流趋势预测区间。再次,为了探究局部多项式回归方法的预测效果,在模拟研究中,将其与LOESS和k近邻回归方法进行比较。最后,以贵阳市每2 min交通卡口车流量数据为例,运用所提方法检测近平稳状态并进行短期交通流趋势预测。结果表明:局部多项式回归方法在可决系数和均方根误差两个评价指标下均表现较优;所提方法对于交通流近平稳状态的检测具有有效性;对于短期断面交通流趋势的预测,预测区间趋势与观测时序基本吻合,且交通流趋势预测区间覆盖率高达80%,进一步说明所提预测方法具有一定的时效性和有效性。
In order to issue road traffic state and traffic flow trend warnings more accurately and in real time,the Pruned Exact Linear Time(PELT)changepoint detection method based on dynamic programming and pruning technology is used for the study on traffic flow state detection and short-term prediction based on bayonet data.First,for the detection of the near-stationary state of traffic flow,the method for detecting near-stationary state of traffic flow is constructed combining with Augmented Dickey-Fuller(ADF)test for time series.This method can quickly and accurately divide the cross-section traffic flow sequence into multiple candidate intervals that may be close to a stationary state,then performs ADF test on each candidate interval,judges whether the candidate interval is stable under the given significance level,and then detects the near-stationary state in the traffic flow sequence.Second,for the short-term prediction of traffic flow,on the basis of the PELT changepoint detection,combining local polynomial regression with the Sigma principle,the short-term section traffic flow trend prediction interval is obtained.Third,in order to explore the prediction effect of the local polynomial regression method,it is compared with the Locally Weighted Regression and Smoothing Scatterplot(LOESS)and k-nearest neighbor regression method in the simulation study.Finally,taking the traffic volume data of the traffic bayonet every 2 minutes in Guiyang City for example,the near-stationary state is detected and the short-term traffic flow trend is predicted by the proposed method.The result shows that(1)the local polynomial regression method performs better under the evaluation indicators of coefficient of determination and root mean square error;(2)the proposed method is effective for detecting the near-stationary traffic flow;(3)for the prediction of short-term traffic flow trend,the prediction interval trend is basically consistent with the observation sequence,and the coverage rate of the traffic flow trend prediction interval is as high as 80%,which further explained that the proposed prediction method has certain timeliness and effectiveness.
作者
陈王勇
胡尧
CHEN Wang-yong;HU Yao(School of Mathematics and Statistics,Guizhou University,Guiyang Guizhou 550025,China;State Key Laboratory of Public Big Data,Guiyang Guizhou 550025,China)
出处
《公路交通科技》
CAS
CSCD
北大核心
2022年第1期120-129,共10页
Journal of Highway and Transportation Research and Development
基金
国家自然科学基金项目(12161016,11661018)
贵州省数据驱动建模学习与优化创新团队项目(黔科合平台人才[2020]5016号)。
关键词
城市交通
交通流短期预测
变点检测
卡口数据
非参数回归
交通状态检测
urban traffic
traffic flow short-term prediction
changepoint detection
bayonet data
nonparametric regression
traffic state detection