Outlier Detection in Categorical, Text and Mixed Attribute Data

Abstract

A significant number of attributes in real data sets are not numerical, but have categorical values. For example, while demographic data may contain quantitative attributes such as the age, many other attributes such as sex and zip code are categorical. Data collected from surveys may often contain responses to multiple-choice questions, which are categorical. Similarly, many kinds of data such as the names of people and entities, IP-addresses and URLs are inherently discrete in nature. In many cases, categorical and numerical data are found in the same data set, as different attributes. This is referred to as mixed-attribute data. Mixed data is quite challenging to address because of the difficulties in appropriately weighting the importance of the different attributes.