基于朴素贝叶斯分类的高速公路交通事件检测

A Naive Bayesian Classifier based Algorithm for Freeway Traffic Incident Detection

摘要:

提出一种基于朴素贝叶斯分类的高速公路非重现交通事件检测算法.将交通事件的检测看作是01分类问题，采用交通波动理论建立交通事件的特征属性概念模型，并利用分段离散化的方法将连续特征变量转换为离散特征变量，设计基于朴素贝叶斯算法的交通事件分类器.以典型高速公路的一条路段进行VISSIM仿真试验.结果表明：该算法的检测率高，且在高强度状况下，算法鲁棒性良好，适用于高速公路交通事件检测系统.

Abstract:

This paper presents a naive Bayesian classifier based algorithm for freeway non recurrent traffic incident detection to enhance the accuracy and learning ability of intelligent traffic incident detection algorithm. The traffic wave theory is employed to establish a conceptual characteristic model of traffic incident, continuous characteristic variables are transferred into discrete characteristic variables via sub discretization, and the naive bayesian based traffic incident classifier is designed by regarding traffic incident detection as “0 1” classification problems. An experiment is carried on a section of a typical freeway, and the performance of the presented model and algorithm is validated via VISSIM simulation. Extensive simulation results show that the algorithm in freeway traffic incident detection system is of high accuracy and strong robustness even if the traffic volumes increase.