Outlier detection is a very important technique in data mining .In this paper, a more practicable system on outlier mining technique in high-dimensional is presented, which uses reduction character of rough set to cut out some inessential attributes and then mines outliers in subspace of every correlation rules. In this density-based outlier mining algorithm, it takes two divided methods to get k- nearest neighbor, which efficiently reduces time complexity and space complexity. As analysis of data shows, this algorithm can find the outliers in high-dimensional space effectively.
|