论文部分内容阅读
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 excavate the trend of traffic flow changes, this paper presents two kinds of methods based on the coarse-grained representation of traffic information: distance-based and binary methods. Based on the overall trend of traffic flow, the distance-based method can set the weighted Euclidean distance of two traffic flow time-varying vectors not to exceed the tolerable threshold. The binary method starts from the individual and considers the traffic state represented by each dimension Consistent, or must be less than the tolerable threshold. The scope of application of the two methods is not only the intersection, but also have their own feature set. The floating car data of a road in Beijing for three months and five months are analyzed and the data are clustered by K-means method. The results show that the traditional K-means can only be clustered from the perspective of pure mathematics and can not completely separate the traffic flow trend. Both the distance-based and the binary methods can further merge the trend of traffic flow and achieve the expected result .