Outlier Detection on Sale Transactions Using Box Plot

Abstract

Outlier detection is an important task in data mining activities and has much attention in both research and application. A value that lies outside which is much smaller or larger than most of the other value in a set of data, this value is called outlier. Most databases include a certain amount of exceptional values, generally termed as outlier. Extremes values tend to be encountered whenever researchers attempt to measure and characterize real world phenomena. Therefore, researchers in all fields are faced with the problem of extreme observations. An observation that is usually large or small relative to the data values is called univariate outlier. In this paper, we present an approach to automating the process of detection univariate outliers. The process is based on graphical display method of construction box plot. In this paper we used outlier labeling method of box plot to detect outlier in electronic items sale database.