摘要
为解决传统轨迹异常检测算法通常综合考虑轨迹所有总体属性,对轨迹在单个属性上的异常检测率不高的问题,提出一种能够有效检测出轨迹单个属性异常的检测算法。通过对轨迹的单个总体属性进行独立处理,利用改进的CanopyKMeans聚类算法对历史数据进行聚类处理,依据聚类的结果建立单个属性的正常行为模式库。利用各属性的正常行为模式库进行对应属性的异常检测,有效提高异常检测的准确率。定义异常程度属性实现对异常行为的分类,检测出异常行为的目的意图,为对异常行为进行针对性处理提供指导。
To solve the problem that the traditional trajectory anomaly detection algorithm usually considers all the global attributes of the trajectory,but its detection rate of the trajectory anomaly on a single attribute is not high,a detection algorithm that could effectively detect the single attribute anomaly of the trajectory was proposed.The individual global attributes of the trajectory were processed independently,and the historical data were clustered using the improved Canopy-KMeans clustering algorithm.The normal behavior pattern database of the individual attributes was established based on the clustering results.The normal behavior pattern database of each attribute was used to detect the corresponding attributes,and the accuracy of anomaly detection was effectively improved.The attribute of abnormal degree was defined to classify the abnormal behavior,and the purpose and intention of the abnormal behavior were detected,to provide guidance for dealing with the abnormal behavior pertinently.
作者
余龙华
张琨
蔡颖
景鸿斐
YU Long-hua;ZHANG Kun;CAI Ying;JING Hong-fei(School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094,China)
出处
《计算机工程与设计》
北大核心
2019年第12期3443-3450,共8页
Computer Engineering and Design
关键词
异常检测
数据挖掘
聚类
异常程度
目的意图
anomaly detection
data mining
clustering
degree of abnormality
purpose intention
