Abstract

An outlier is a data instance that is dissimilar to other instances in a dataset. The goal of outlier analysis methods is, in the first plane, to detect those instances that are the most dissimilar. We will show how this concept needs to be modified when knowing that a dataset consists of two or more classes, and that these knowledge is important for data analysis. (contains a class attribute). We describe a concept of class-based outlier detection that aims to detect outliers that could not be found without knowing their class membership. We will present several areas where such outlier detection is important, and also introduce methods for detecting class-based outliers.

Short Bio:

Luboš Popelínský is an Associate professor at the Department of Machine Learning and Data Processing & Chair of Knowledge Discovery Laboratory of the Faculty of Informatics, Masaryk University, Brno, Czech Republic. He teaches machine learning, text mining, and logic courses.

About Talks@DCC 

The mission of the Talks@DCC seminars is to bring together researchers and students, to foster discussions and promote scientific awareness and collaboration.

Participate: attend and propose your talk.