Data Mining: Outlier Detection

doi:10.1016/B978-0-12-809633-8.20386-5

Reference Module in Life Sciences

Encyclopedia of Bioinformatics and Computational Biology

Volume 1, 2019, Pages 456-462

https://doi.org/10.1016/B978-0-12-809633-8.20386-5 Get rights and content

Abstract

Outlier detection techniques aim at singling out the observations that most deviate from the rest of the data at hand. During the years different notions of anomaly have been provided, each one with its peculiarities. Unsupervised methods search for outliers in an unlabelled dataset by assigning to each object a score which reflects its degree of abnormality. In this paper, we provide an overview of unsupervised data mining outlier detection techniques by considering different families of approaches, including statistical-based, distance-based, density-based, isolation-based, angle-based, subspace-based, ensembles, and methods providing explanations. Advantages, drawbacks and potential applicative scenarios are also highlighted.

References (0)

Cited by (0)

Fabrizio Angiulli received the Laurea degree in computer engineering in 1999 from the University of Calabria (UNICAL). He is an associate professor of computer engineering since 2011 at DIMES Department, University of Calabria, Italy. In 2013, he obtained the Italian Full Professor qualification. Previously, he held a research and development position at ICAR of the National Research Council of Italy and, after that, a tenured assistant professor position at DEIS, University of Calabria. His main research interests are in the area of data mining, notably outlier detection and classification techniques, knowledge representation and reasoning and database management and theory. He has authored more than 80 papers appearing in premier journals and conference proceedings. He regularly serves on the program committee of several conferences and, as an associate editor, on the editorial board of the AI Communications journal. He is a senior member of the IEEE.

View full text