小样本道路旅行时间数据中的异常点剔除算法

 刘喜梅， 刘义芳， 高林

（青岛科技大学 自动化与电子工程学院，山东 青岛 266042）

 摘要: 利用车牌照匹配技术获取的小样本旅行时间数据中通常夹杂大量异常点，无法直接用以表征当前交通状态及交通旅行时间数据的动态、离散、小样本等特性，在传统剔除算法的基础上，提出了一种统计分析与模糊C均值聚类相结合的异常点剔除新方法。将新剔除方法与传统剔除方式效果进行分析比较，得出一种精确度较高的异常点剔除方法。仿真结果表明，该方法在处理交通小样本数据上，大幅度提高了异常点检测的准确性，能够有效过滤异常数据。

关键词: 智能交通； 旅行时间； 统计分析； 模糊C均值聚类； 异常点剔除

中图分类号： U 491.2+65                 文献标志码： A

Algorithm Outlier Filtering for Small Simple Data of Travel Time

LIU Ximei, LIU Yifang, GAO Lin

(College of Automation and Electronic Engineering, Qingdao University of Science and Technology, Qingdao 266042, China)

Abstract: As there are usually many outliers while using the method of license plate matching to obtain travel times, traffic data with small size can′t be used directly. In order to deal with such common problems, we have studied the traditional approaches about filtering outliers. Then a new outlier filtering method, which based on the features of the traffic data and combined the statistical analysis with the fuzzy Cmeans clustering, was proposed. After comparing the new way with the traditional method, we obtained a method with higher degree of accuracy in filtering outliers. The results revealed that the new method can detect the outliers exactly and eliminate the outliers efficiently for small sample data.

Key words: intelligent transportation; travel time; statistical analysis; fuzzy Cmeans clustering; outlier filtering

 收稿日期： 20140401

作者简介： 刘喜梅（1961—），女，教授，博士生导师.

文章编号： 16726987（2015）03034604； DOI： 10.16351/j.16726987.2015.03.022