离群数据的探测

综述了离群数据(outliers)探测是数据挖掘和知识发现的一项重要任务及其在天学中兴起的必然性。简要介绍了离群数据的定义、特点、产生原因及影响，着重阐述了探测一维离群数据和多维离群数据的方法，并且与一些聚类算法作了对比。每一种算法各有优劣，天学家应根据天数据的特点，探讨出适合天数据特点的离群数据探测方法，以发现一些不同寻常的、稀有的，甚至新类型的天体和天现象。

The Outlier Detection

The outlier detection is an important task of data mining and knowledge discovery in database. Its necessities in astronomy are reviewed with the definitions, characteristics, forming reasons and causing effects of outliers being simply introduced. The methodologies to detect univariate outliers and multivari-ate outliers are summarized, and compared with other clustering algorithms. Since each kind of approach has its own pros and cons, astronomers should reasonably choose methodologies to detect outliers according to the characteristics of astronomical data, so that some unusual, rare or even new astronomical objects and phenomena could be found.