Outlier Detection Based on Boundary and Distance
Cungen Cao · Dianzi xuebao · 2010
In recent years,outlier detection has gained considerable interest.The identification of outliers is important for many applications such as intrusion detection,credit card fraud,criminal activities in electronic commerce,medical diagnosis and anti-terrorism,etc.The aim of outlier detection is to find small groups of objects who behave in an unexpected way or have abnormal properties when compared with the rest large amount of data.Since the existing methods for outlier detection cannot deal with uncertain and incomplete data.In this paper,we propose a new method for outlier definition and detection,which exploits the basic notion —boundary of rough sets and Knorr's method about distance-based outliers.We also give an algorithm BDOD to find such outliers within the framework of rough set theory.The effectiveness of our algorithm is demonstrated on publicly clinical diagnosis data sets.Our method gives a new approach to the treatment of uncertain and incomplete data in outlier detection.