基于AdaBoost分类器的实时交通事故预测

发布时间：2018-06-09 06:10

本文选题：智能交通 + 事故预测　；参考：《计算机应用》2017年01期

【摘要】：传统的道路交通事故预测是对交通事故次数及其造成的损失的历史趋势进行预测,针对其不能反映交通事故与实时交通特性关系、不能有效地预防事故发生的问题,提出一种基于AdaBoost分类器的交通事故实时预测的方法。首先,将交通道路划分为正常、危险两种交通状态,利用实时采集的交通流数据作为特征变量对不同的状态进行表征,将事故的实时预测问题转化为分类问题;然后,采用Parzen窗非参数估计的方法对两种状态在不同时间尺度下候选交通流特征的概率密度函数(PDF)进行估计,利用基于概率分布的可分性判据分析估计的密度函数,选择合适的特征变量及时间尺度,确定样本数据;最后,根据样本数据训练AdaBoost分类器对不同的交通状态进行分类识别。实验结果表明,采用交通流特性的标准差特征对测试样本分类的正确率比平均值特征高7.9%,更能反映不同交通状态的差别,获得更好的分类结果。
[Abstract]:The traditional prediction of road traffic accidents is to predict the historical trend of the number of traffic accidents and the losses caused by them. It can not reflect the relationship between traffic accidents and real-time traffic characteristics, and can not effectively prevent the occurrence of accidents. A method of real-time traffic accident prediction based on AdaBoost classifier is proposed. First of all, the traffic road is divided into two traffic states: normal and dangerous. The traffic flow data collected in real time are used as characteristic variables to represent the different states, and the real-time prediction of accidents is transformed into classification. The probability density function (PDF) of the candidate traffic flow characteristics of two states at different time scales is estimated by using the Parzen window nonparametric estimation method. The estimated density function is analyzed by using the divisibility criterion based on the probability distribution. The appropriate feature variables and time scales are selected to determine the sample data. Finally, the AdaBoost classifier is trained to classify different traffic states according to the sample data. The experimental results show that the correct rate of classification of test samples by using the standard deviation features of traffic flow characteristics is 7.9 higher than that of average features, which can reflect the differences of different traffic states and obtain better classification results.
【作者单位】：天津大学电气与自动化工程学院;
【分类号】：U491.31;O213

【相似文献】