Outlier detection and evaluation by network flow

• DOI: 10.1504/ijcat.2008.021946
• Publication Date: 2008-12-11T07:38:12Z
• Authors: Ying Liu; Alan P. Sprague
• Keywords: Local outlier factor; Data set

This paper introduces a novel method to separate abnormal points from normal data, based on network flow. approach uses the Maximum Flow Minimum Cut theorem graph theory find outliers and strong outlier groups, evaluate by degrees. Similar are discovered together delivered user together; in an application where of greatest interest, this will allow similar be analysed together. Effectiveness is demonstrated comparison with three other detection algorithms. Further experimental testifies algorithm can improve query accuracy content-based image data set. effective higher dimensional as well low dimension.