摘要
针对城市道路交通系统中交通流演变的随机性和复杂性,以及实时交通状态判别本身所具有的不确定性,基于模式识别和相似预报的思想,提出了一种交通状态概率预报的K近邻非参数回归模型。模型首先利用城市道路路段上环形线圈采集的交通流数据,采用模糊聚类技术,生成历史样本数据库;接着采用相似离度指标进行近邻搜索;然后根据近邻子集,构建交通状态概率预报函数,对路段未来时段的交通流运行状态进行预报,并用概率定量描述该状态发生的可能性大小。最后根据该模型,结合实际数据,进行了不同预报时长的分级交通状态的概率预报试验,独立样本检验结果表明,该模型预报准确率高,稳定性好。
According to the stochastic property and complexity of the traffic flow evolution on the urban road and the uncertainty of traffic state discrimination,based on the thoughts of pattern recognition and similar forecasting,a model of K-nearest neighbor nonparametric regression for traffic state probability forecasting was put forward.First,Fuzzy clustering was employed to generate the historical sample database using the traffic flow data from loop detector on the urban road section.Then the index of analogy deviation was used to search the nearest neighbors,based on which the function of traffic state probability forecast is constructed to predicted the traffic state of the next time period,and the possibility of the coming state was quantitatively described using probability.Finally,according to the proposed model,the graded traffic state probability forecasting test of different time periods was carried out using the field traffic flow data.The result of independent sample test indicates that the model has a finer precision and stability.
出处
《公路交通科技》
CAS
CSCD
北大核心
2010年第8期76-80,共5页
Journal of Highway and Transportation Research and Development
基金
国家自然科学基金重点项目资助(70631002)
国家高技术研究发展计划(八六三计划)资助项目(2008AA11Z205)
关键词
交通工程
概率预报
非参数回归
交通状态
相似预报
traffic engineering
probability forecasting
nonparametric regression
traffic state
similar forecasting
