摘要
船舶自动识别系统所产生的数据巨量且复杂,异常检测作为航行数据挖掘的重要部分,近年来被许多学者所研究,为海上监管部门检测和分析船舶异常行为提供了有力的数据支持。为检测船舶航行过程中产生的少量且异常的数据,采用了一种基于iForest的检测算法对船舶轨迹点异常进行了研究,对算法过程的孤立树和子采样的个数提出一种自适应选择方法,根据不同数据集算法可以提取出最适宜的参数,从而优化检测结果。利用渤海部分区域的AIS数据进行实验研究,并将结果与三种常用的异常检测算法进行了比较,实验结果表明,该方法的AUC数值高于其他算法。
The data generated by the automatic ship identification system is huge and complex. Anomaly detection, as an important part of navigation data mining, has been studied by many scholars in recent years,providing strong data support for maritime regulatory authorities to detect and analyze abnormal ship behavior.In order to detect a small amount of abnormal data generated during the navigation of ships, this paper proposes an anomaly detection algorithm for ship trajectory points based on iForest. The algorithm can use random hyperplane segmentation data sets according to different dimensions of AIS data, which simplifies Distance calculation between multidimensional data. On this basis, an adaptive parameter selection method is proposed, which can extract the most suitable parameters according to different data set algorithms, so as to optimize the detection results. The AIS data in part of the Bohai Sea is used for experimental research, and the results are compared with three commonly used anomaly detection algorithms. The experimental results show that this method has a better effect on detecting multi-dimensional ship trajectory point data.
作者
梁超凡
刘欣钰
Liang Chaofan;Liu Xinyu(Liaoning Key Lab of Physical Geography and Geomatics,Liaoning Normal University,Dalian,China;School of Geographical Sciences,Liaoning Normal University,Dalian,China)
出处
《科学技术创新》
2023年第3期96-99,共4页
Scientific and Technological Innovation
