交通运输系统工程与信息 ›› 2014, Vol. 14 ›› Issue (2): 87-93.

• 系统工程理论与方法 • 上一篇    下一篇

基于K-means聚类算法的公交运营时段分析

沈吟东*,张仝辉,徐 甲   

  1. 华中科技大学自动化学院,图像信息处理与智能控制教育部重点实验室,武汉430074
  • 收稿日期:2013-08-30 修回日期:2013-12-04 出版日期:2014-04-25 发布日期:2014-07-07
  • 作者简介:沈吟东(1965-),女,安徽合肥人,教授,博士.
  • 基金资助:

    国家自然科学基金(71171087,70971044);国家社会科学基金重点项目(13&ZD175)

Homogeneous Bus Running Time Bands Analysis Based onK-means Algorithms

SHEN Yin-dong,ZHANG Tong-hui,Xu Jia   

  1. Image Processing and Intelligent Control Key Laboratory of Education Ministry, School of Automation, Huazhong University of Science and Technology, Wuhan430074, China
  • Received:2013-08-30 Revised:2013-12-04 Online:2014-04-25 Published:2014-07-07

摘要:

公交车辆在高低峰等不同时段的运营时间差异较大,因此,只有对各时段的运 营时间分别加以分析才能准确掌握运营时间规律,这对提高公交运营方案的准点率具有 重要影响,是公交运营分析和优化调度等工作的不可或缺的重要基础.目前我国公交时段 划分主要依据人工经验,简单且粗糙.本文基于大量GPS运营数据,创新性地将K-means 聚类算法应用于运营时段划分,并结合公交样本数据特点,提出一种改进的K-means聚 类算法,其中改进了传统的初始簇中心选择方法,并设计了利用三角形不等式减少不必 要的距离计算和基于模糊聚类思想的簇中心更新算法.十堰市和海口市公交的案例分析 表明,本文的K-means聚类方法可行,改进算法的计算效率更高,划分的时段与实际调研 分析结果更加吻合.

关键词: 智能公交, 时段划分, K-means聚类算法, 运营分析, 数据挖掘

Abstract:

Bus running time normally varies significantly by different time period like peack and non-peak hours. Therefore, the running time should be measured for individual time period, which affects the on-time probability of schedules. Setting precise homogeneous running time (HRT) bands is essential for service reliability measurement and scheduling. In China, the HRT bands are manually set based on experiences, and the HRT bands generated lack accuracy. With GPS data, this paper uses theK-means clustering algorithm to divide HRT bands. Then, theK-means algorithm is improved and an enhanced HRT bands division method is developed. Several methods are also addressed: an enhanced cluster initialization method, a cluster center updating method based on triangle inequality, and fuzzy clustering method. The field study on Shiyan Bus and Haikou Bus demonstrates the feasibility of the clustering method.The generated HRT bands match well with real-world situations.

Key words: intelligent transportation, homogeneous running time band, K- means algorithm, bus service measurement, data mining

中图分类号: