Modeling and efficient mining of intentional knowledge of outliers

In this paper, we study in a general setting the notion of outliered patterns as intentional knowledge of outliers and algorithms to mine those patterns. Our contributions consist of a model for defining outliered patterns with the help of categorical and behavioral similarities of outliers, and efficient algorithms for mining knowledge sets of distance-based outliers and outliered patterns. Our algorithms require only very limited domain knowledge, and no classified information. We also present an empirical study to show the feasibility of our algorithms.