基于OBD数据的驾驶人出行模式挖掘

Drivers'Travel Pattern Mining Based on OBD Data

传统的出行模式研究通常依靠问卷调查分析驾驶人出行特征,所得结果易受调查数据主观性影响,针对此问题基于北京市域范围内2个月共计3570辆私家车的车载诊断数据,对驾驶人的不同出行模式进行分析并建模。通过长期采集的车辆各项参数,采用基于密度峰值的聚类算法进行聚类,将不同的驾驶人分为高频出行者、通勤出行者、长距偶发出行者以及危险出行者,并从平均出行距离、出行频次、百公里危险驾驶行为次数和出行时段等多维度进行分析,反映驾驶人行为的变化性和规律性。根据聚类的结果,使用多维离散隐马尔可夫模型进行建模并完成测试。测试表明,所提出的算法对于驾驶人出行模式的识别具有较高的准确性,对于4种类型的出行者,平均识别率超过91%,最高识别率可达94.5%。

The traditional travel pattern research mainly relies on questionnaires to analyze the driver's travel charac⁃teristics,the result of which is not objective.In order to solve the problem,the study analyzed and identifieddifferent⁃drivers'travel patterns based on the vehicle on-board diagnosticdata from 3570 private cars in Beijing within two months.According to the parameters recorded from vehicles,a clustering algorithm called Clustering by Fast Search and Find of Density Peaks was used to classify different drivers into high-frequency travelers,commuting travelers,long-distance and occasional travelers and dangerous travelers,and analyzed from the aspects of average travel dis⁃tance,travel frequency,travel time and dangerous driving behavior times of 100 km,to reflect the variability and reg⁃ularity of driver's travel pattern.According to the clustering result,the multi-dimensional discrete Hidden Markov Model was used for modeling and measurement.Results indicate that the algorithm proposed shows good accuracy on the identification of drivers'travel patterns.For different kinds of drivers,the averagecorrect recognition rate exceed 91%while the highest recognition rete can reach 94.5%.