基于时空大数据的异常车辆检测

Abnormal vehicle detection based on spatio-temporal big data

针对传统人工检查黑车的方式不但耗时耗力而且效率低下的问题,提出一种新的自动检测黑车的方法。在Hadoop平台上,对物联网技术采集的全疆车辆加气数据进行分析;抽取车辆加气的时间特征和空间特征;利用随机森林算法研究车辆与驾驶员、加气站间的关系,从而发现具有异常加气模式的黑车车辆。在大规模真实数据集上的实验表明:提出的方法在黑车发现问题上有较高的准确率,可以用于帮助有关部门提高黑车检测的效率。

Aiming at the problem that traditional method of unlicensed taxis manual check is not only timeconsuming but also inefficient,a novel automatic unlicensed taxis detection method is proposed. Vehicles refueling data of Xinjiang Province collected by Internet of things( IOT) are analyzed on Hadoop platform. Spatial features and temporal features of vehicles refueling are extracted. Random forest algorithm is applied in research on relationship between vehicles,drivers and gas stations to discover unlicensed taxis with abnormal refueling patterns. Experiments on real-world large-scale dataset show that the proposed method has high accuracy in detecting unlicensed taxis,and its application can help relevant department improve the efficiency of unlicensed taxis detection.