摘要
为挖掘典型的交通流变化趋势,本文结合交通信息的粗粒度表达方式,提出了2种基于浮动车数据的提取方法:定距型方法和二值型方法。定距型方法从交通流整体趋势出发,设定2个交通流时变向量的加权欧几里得距离不能大于可容忍阈值;二值型方法从个体出发,考虑每一维度表示的交通状态必须一致,否则必须小于可容忍阈值。2种方法的适用范围既有交集,也有各自的特征集。对北京一条路段3个月周五的浮动车数据进行分析,并采用K-均值法对数据进行初步聚类。分析结果表明,传统的K-均值法只能从纯数学角度聚类,不能将交通流趋势完全区分开;而定距型和二值型方法均能够将交通流变化趋势进一步合并,达到预期效果。
In order to mine typical traffic flow trend, based on the coarse granularity expression of traffic information, we presented two extraction methods based on floating car data： distance type and binary type. Distance type method considers whole traffic flow trend, and defines that the Euclidean distance between two traffic flow time-dependent vectors must be less than threshold that can tolerance. Binary type method defines that the traffic state of each dimension must be consistent from the point of individual, or the difference of each dimension should be less than the bearable threshold. The two methods not only have intersection set, but also have their own characteristic set. Taking one link of Beijing for example, the raw data is the floating car data of Friday in three months, and K-Means is used to preliminary cluster the raw data. The analysis result shows that （1） traditional K-means can only cluster data from the mathematic perspective, and some traffic trends cannot be entirely separated; （2） the distance type and binary type methods can further merge these traffic trends and achieve the desire result.
出处
《公路交通科技》
CAS
CSCD
北大核心
2013年第12期125-131,共7页
Journal of Highway and Transportation Research and Development
基金
国家高技术研究发展计划(八六三计划)项目(2012AA12A207)
关键词
交通工程
交通流变化趋势
聚类分析
浮动车数据
交通状态
traffic engineering
traffic flow trend
cluster analysis
floating car data
traffic state
