A Survey for Outlier Detection and its Strategies

Abstract

Outlier detection is the most important research problem in data mining that aims to detect outliers from high volumes of data. The Outlier detection problem has sophisticated applications in the field of Fraud detection for Credit cards, Military supervision for enemy activities, E-mail spam detection etc. Most such applications are high dimensional domains in which the data can contain hundreds of dimensions . Most approaches use the concept of proximity in order to find outliers based on relationship to the rest of the data. But it fails when data comes with high dimensions. In order to find out those outliers, we introduce a survey of sophisticated techniques for outlier detection. In this paper, we identified a well defined mechanisms to handle outliers, their motivations and distinguish them.

Keywords: Outlier detection, Proximity, Data mining, High Dimensionality, Fraud detection.