基于隐马尔科夫模型的片区交通拥堵形态预测  

Prediction of Regional Traffic Congestion Patterns Based on Hidden Markov Model

作　　者：孙婷婷 黄正锋[1] 朱鸿东 郑彭军[1] SUN Ting-ting;HUANG Zheng-feng;ZHU Hong-dong;ZHENG Peng-jun(Faculty of Maritime and Transportation,Ningbo University,Ningbo 315832,China)

机构地区：[1]宁波大学海运学院,浙江宁波315832

出　　处：《数学的实践与认识》2020年第21期98-108,共11页Mathematics in Practice and Theory

基　　金：浙江省自然科学基金(LY18E080009);宁波市自然科学基金(2019A610040)

摘　　要：若城市片区内多条相连路段同时堵塞,容易引发道路网络瘫痪,有必要对片区交通拥堵形态进行预测,为主动交通管理提供信息支撑.由于道路上下游的交通流量联系'紧密,因此通过隐马尔可夫模型建立片区外围与内部交通状态的关联,用于片区拥堵形态预测.首先,清洗和挖掘大量浮动车轨迹数据,获得道路交通状态信息.然后,将外围交通状态和内部片区拥堵形态分别作为观察和隐藏状态,构建隐马尔可夫预测模型.最后,以宁波主城某医院所在片区为例对模型的有效合理性进行验证,预测准确率可以达到83.4%,与自回归滑动平均模型相比,高出6.2%.If multiple connected road links in a small-scale zone are blocked at the same time,it will easily cause the road network to be paralyzed,so it is necessary to predict the traffic congestion patterns in the block to provide information for active traffic management.Due to the close connection between the upstream and downstream traffic flow of the road,the hidden Markov model is used to establish the relationship between the external traffic states and internal congestion states of the small-scale zone and to predict the congestion patterns of the small-scale zone.Firstly,a large number of floating vehicle trajectory data are cleaned and mined to obtain road traffic status information.Then,the external traffic state and the internal congestion state are regarded as observation and hidden state respectively,and the hidden Markov prediction model is constructed.Finally,the validity and rationality of the model is verified by taking a hospital zone in Ningbo as an example.The prediction accuracy can reach 83.4%,which is 6.2% higher than that of Autoregressive Moving Average model.

关 键 词：片区拥堵 隐马尔科夫模型 拥堵形态 

分 类 号：U491.265[交通运输工程—交通运输规划与管理]