文摘
In data mining, outlier analysis is an important task. In recent years, many techniques have been developed for outliers mining. This paper provides an overview of the existing methods for outlier analysis. And on the base of these researches, we first propose the definition of local outlier and some formulas, and then develop a clusteringbased approach to mining local outliers. The clustering algorithm used in this paper is based on the maximal frequent itemsets in association rule mining, the properties of maximal frequent itemsets are applied to clustering. After clustering, the objects which have many common features are grouped together. By calculating the standard deviation of a cluster with respect to a certain attribute and local outlier factor of an object with respect to a certain attribute, local outliers are identified. The experimental results show that our approach is effective and appropriate to mining local outliers.