首页> 外文会议>International conference on communications and cyber physical engineering >Improved Method for Noise Detection by DBSCAN and Angle Based Outlier Factor in High Dimensional Datasets
【24h】

Improved Method for Noise Detection by DBSCAN and Angle Based Outlier Factor in High Dimensional Datasets

机译:高维数据集中基于DBSCAN和基于角度的离群因子的噪声检测的改进方法

获取原文

摘要

Various data mining methods are used to detect outliers from different databases. It is essential to detect outliers in different kinds of real time application areas, such as the health care sector and insurance, marketing, banks and finance et c. The proposed method in this paper is a new outlier detection method from a high dimensional data sets which combines the angle based outlier detection (ABOD) with the classical density based clustering method DBSCAN. The algorithm consists of three stages in which the first stage consists of applying of the PCA on the data set which will result in a subset of attributes, to this subset of attributes the dbscan algorithm is applied which results in detection of a set of outliers. In the third and final stage ABOD is applied to the set of outliers. Experimental analysis conducted state that the result improves the detection accuracy and decreases the number of false positives.
机译:各种数据挖掘方法用于检测来自不同数据库的异常值。必须检测不同种类的实时应用领域中的异常值,例如医疗保健部门和保险,市场营销,银行和金融等。本文提出的方法是一种新的高维数据集离群值检测方法,它将基于角度的离群值检测(ABOD)与基于经典密度的聚类方法DBSCAN相结合。该算法由三个阶段组成,其中第一阶段包括在数据集上应用PCA,这将导致属性的子集,对dbscan算法应用到该属性子集的结果将检测到一组异常值。在第三阶段也是最后阶段,ABOD应用于异常值集。实验分析表明,该结果提高了检测精度,并减少了误报的数量。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号