Abstract:

Outlier detection has great significance in data processing. At present, its detection methods can be roughly divided into statistical based, distance based, density based and clustering based. In order to grasp the current research status of outlier detection methods based on clustering technology, this paper introduces and classifies the representative outlier detection methods based on clustering, which are mainly divided into four categories：detection methods in static data sets, detection methods in data streams, detection methods in large-scale data and other methods. The problems, algorithm ideas, application scenarios and their advantages and disadvantages of each method are summarized and analyzed in detail, and the existing problems are analyzed and the future development direction is proposed.

Key words: outlier detection, clustering, static data, data stream, large-scale data set

摘要：

离群点检测在数据处理中具有重要研究意义，其检测方法大致可以分为基于统计、基于距离、基于密度和基于聚类的方法。为了及时掌握当前基于聚类技术的离群点检测方法的研究现状，通过归纳与整理，将具有代表性的基于聚类的离群点检测方法进行了介绍和归类，将其主要分为静态数据集中的检测方法、数据流中的检测方法、大规模数据中的检测方法和其他方法等四大类。对每类方法所解决的问题、算法思想、应用场景以及各自的优缺点进行了详细的归纳和分析，指出目前存在的问题以及未来发展方向。

关键词: 离群点检测, 聚类, 静态数据集, 数据流, 大规模数据集