关键词: 离群点检测； k-means聚类； 局部离群因子

Abstract:

Most existing outlier detection algorithms ignore local information of data sets, they are of low accuracy. We adopted a twophase algorithm based on k-means clustering algorithm, defined a new local stray factor as the standard to judge whether data objects are outliers. We also improved the process of detecting outliers and solved the above problem. Experiments show that our algorithm overcomes the shortcomings of existing methods, ensure the algorithm has linear time complexity and is able to find outliers in data sets more accurately and effectively.

Key words: outlier detecting, k-means clustering, local outlier factor