Data Mining Techniques for Outlier Detection

Abstract

Among the growing number of data mining techniques in various application areas, outlier detection has gained importance in recent times. Detecting the objects in a data set with unusual properties is important as such outlier objects often contain useful information on abnormal behavior of the system described by the data set. Outlier detection has been popularly used for detection of anomalies in computer networks, fraud detection and such applications. Though a number of research efforts address the problem of detecting outliers in data sets, there are still many challenges faced by the research community in terms of identifying a suitable technique for addressing specific applications of interest. These challenges are primarily due to the large volume of high dimensional data associated with most data mining applications and also due to the performance requirements. This chapter highlights some of the important research issues that determine the nature of the outlier detection algorithm required for a typical data mining application. The research issues discussed include the method of outlier detection, size and dimensionality of the data set, and nature of the target application. Thus this chapter attempts to cover the challenges and possible research directions along with a survey of various data mining techniques dealing with the outlier detection problem.