TRAFFIC INCIDENT DETECTION BASED ON ROUGH SETS APPROACH

摘要：

This paper presents an approach to detect traffic incident which uses the rules generated by rough sets theory to classify traffic patterns for incident detection.Performance metrics such as detection rate, false alarm rate, mean time to detection and classification rate are computed.By way of illustration, a simulated traffic data set which is balanced and the real 1-880 freeway traffic data collected in California which is imbalanced are used to assess the detection performance of this approach.Rough sets method is sensitive to attributes discretization proven by the experimental results, so cross validation was used to conduct the discrete operation in order to improve the classification accuracy.Further tests also indicate that rules filter can enhance the performance of classification.Our experiments illustrate the incident detection models based on rough sets theory have favorable performance compared with those based on support vector machine.At last, a brief conclusion as well as future research needed is also discussed.

关键词： Rough sets theory Automatic incident detection Support vector machine Cross validation

作者: SHU-YAN CHEN WEI WANG GAO-FENG QU

作者单位: Optoelectronics key laboratory of Jiangsu province, Nanjing Normal University, Nanjing 210097, China College of Transportation, Southeast University, Nanjing 210096, China College of Automation, Nanjing University of Posts and Telecommunications, Nanjing 210003, China

会议类型: 国际会议

会议名称: 2007 International Conference on Machine Learning and Cybernetics(IEEE第六届机器学习与控制论国际会议)

会议地点: 香港

会议语种:英文

页码: 3734-3739

在线出版日期: 2007-08-19（万方平台首次上网日期，不代表论文的发表时间）