城市交叉口群的聚类分析分类算法

Clustering Analysis and Classification Algorithm for Urban Intersections

基于交叉口流量数据的交叉口分类算法,较好地兼顾了协调控制和个性化配时方案两方面,同一类的交叉口配时方案相似,地理位置或距离相近(易于划分子区)。本研究首先使用多重评估指标法和组内平方误差和的方法,确定聚类分析的分类个数。然后通过使用划分聚类分析中围绕中心点的划分算法(PAM),以交叉口的流量为依据,对苏州工业园区的107个交叉口进行分类,通过建立相异性函数,衡量聚类中交叉口的相似性。将苏州工业园区的107个交叉口分为若干类,每一类的中心点能够体现本类中各点的集体特点,是本类各项特征的综合体现。最终将苏州工业园区107个交叉口分为3类,使得每一类都对族群内各个交叉口具有较高的相似性,对族群外的交叉口差别较大。每一族群的中心点交叉口对本组群包含的交叉口具有良好的代表性,集中体现了本族群交叉口的特征。在107个交叉口中有80个交叉口的总停车时间减少,占到了所有交叉口的74.8%,所有交叉口一共减少了241.325 h的停车时间,每个交叉口平均减少2.26 h。

The intersection classification algorithm based on intersection traffic volume data better considers both coordinated control and personalized timing scheme. The timing schemes for same type of intersection are similar, and their geographical locations or distances are similar(easy to divide sub-zones).First, the multiple evaluation index method and the method of square error sum within the group are used to determine the class number in cluster analysis. Then, by using the partitioning algorithm(PAM) around the center point in the partitioned cluster analysis, based on the traffic volume at intersection, 107 intersections in the Suzhou Industrial Park are classified, and the similarity of the intersections in the cluster are measured by the established dissimilarity function. The 107 intersections of Suzhou Industrial Park are divided into several categories. The center point of each category can reflect the collective characteristics of each point in this category, which is a comprehensive reflection of the characteristics of this category. The 107 intersections in the Suzhou Industrial Park are divided into 3 categories, making each category has a high similarity to the intersections within the community, while has a significant difference to the intersections outside the community. The center point intersection of each community can well represent the intersections included in its community, which fully embodies the characteristics of the intersection of this community. Among the 107 intersections, the total parking time at 80 intersections decreased, which accounts for 74.8% of all intersections. A total of 241.325 h of parking time are reduced at all intersections, and the average parking time at the intersections is reduced by 2.26 h.