Outlier detection based on local minima density
Outlier is great concern in machine learning task and the traditional methods based on nearest neighbor outlier detection have some weaknesses: performance is sensitive to parameter k,and interpretability is not strong.In this paper,based on the novel idea that outlier objects have lower density than their neighbors and relatively large distance from objects with higher density,we propose a new algorithm for outlier detection to overcome the weakness above.We compared the proposed method with other existing methods based on various types of synthetic datasets.We also applied the proposed method in real water quality data.The results of the numerical experiments indicated that the proposed method has better effectiveness,stability,and interpretability on the detection of low-density outlier detection.
outlier local minima density Nearest neighbors
Jia Liu Guoyin Wang
Chongqing Key Laboratory of Computational Intelligence Chongqing University of Posts and Telecommuni Institute of Electronic Information Technology Chongqing Institute of Green and Intelligent Technolo
国际会议
重庆
英文
718-723
2016-03-20（万方平台首次上网日期，不代表论文的发表时间）