摘要
为了精准检测伪周期数据流异常,提出一种基于局部离群因子的伪周期数据流异常检测方法。通过伪周期数据流的空间维度,展开网格划分处理。对邻域粗糙集模型展开分析构建特征冗余度度量特征冗余图,采用图割理论得到特征划分子集。引入聚类簇数评估方法确定最优特征分割,同时根据建立最优类簇结构评估指标展开伪周期数据流特征选择。利用改进孤立森林(iForest)和局部离群因子相结合的方式分别展开周期数据流异常检测,将获取的检测结果融合进而得到最终的伪周期数据流异常检测结果。实验结果表明,所提方法可以有效检测出伪周期数据流异常,获取良好的检测效果。
In order to accurately detect the anomalies of pseudo-periodic data streams,a pseudo-periodic data stream anomaly detection method based on local outliers is proposed.Through the spatial dimension of the pseudo-periodic data stream,the grid division process is carried out.The neighborhood rough set model is analyzed to construct a feature redundancy map with feature redundancy measure,and a subset of features is obtained by using graph cut theory.The optimal feature partitioning is determined by introducing the evaluation method of the number of clusters,and the pseudo-periodic data stream feature selection is carried out according to the establishment of the evaluation index of the optimal class cluster structure.The combination of improved isolated forest(iForest)and local outliers is used to detect the anomalies of periodic data streams,and the results are fused to obtain the final pseudo-periodic data stream anomaly detection results.The experimental results show that the proposed method can effectively detect pseudo-periodic data stream anomalies and obtain good detection results.
作者
陈玉姝
王丽楠
王晨华
CHEN Yu-shu;WANG Li-nan;WANG Chen-hua(College of Computer Science and Electronic Engineering,Xinjiang College of Science&Technology,Korla Xinjiang 841000,China)
出处
《计算机仿真》
2025年第4期386-390,共5页
Computer Simulation
基金
教育部产学合作协同育人项目(230902405254907)
新疆科技学院校级基金项目(JGPT-23-06)。
关键词
局部离群因子
伪周期
数据流
异常检测
Local outlier
Pseudo-periodic
Data streaming
Anomaly detection
