摘要
离群检测是数据挖掘领域的一个重要内容,它为分析各种海量、复杂、含有噪声的数据提供了新的方法.对离群簇进行了定义并据此提出一种离群检测方法,该方法增量式地对原始数据集进行聚类,在得到的簇中寻找离群簇.根据提出的簇间差异性度量,新方法可处理混合属性数据集.同时探讨了参数取值.基于人工数据集和真实数据集上的实验表明,新方法检测离群点具有精度高、速度快的优点,适用于大规模数据集.
Outlier detection is an important branch in data mining field. It provides new methods for analyzing all kinds of massive, complex data with noise. In this paper, an outlier detection algorithm is presented by introducing and discussing the concept of outlier cluster. The algorithm firstly partitions the dataset into several clusters by the incremental clustering approach. Outliers are then detected from the cluster set. Moreover, by introducing inter-cluster dissimilarity measure, the proposed algorithm gains a good performance on the mixed data. At the same time the parameter values are discussed. The experimental results on the synthetic and real-life datasets show our approach outperform the existing methods on identifying meaningful and interesting outliers.
出处
《小型微型计算机系统》
CSCD
北大核心
2010年第11期2282-2286,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60673191
10871040)资助
河南省教育厅自然科学基础研究计划项目(2010A520033)资助
关键词
离群检测
混合属性
离群簇
outlier detection
mixed attributes
abnormal cluster
