Leave-one-out Kernel Density Estimates for Outlier Detection

Sevvandi Kandanaarachchi () and Rob Hyndman

No 2/21, Monash Econometrics and Business Statistics Working Papers from Monash University, Department of Econometrics and Business Statistics

Abstract: This paper introduces lookout, a new approach to detect outliers using leave-one-out kernel density estimates and extreme value theory. Outlier detection methods that use kernel density estimates generally employ a user defined parameter to determine the bandwidth. Lookout uses persistent homology to construct a bandwidth suitable for outlier detection without any user input. We demonstrate the effectiveness of lookout on an extensive data repository by comparing its performance with other outlier detection methods based on extreme value theory. Furthermore, we introduce outlier persistence, a useful concept that explores the birth and the cessation of outliers with changing bandwidth and significance levels. The R package lookout implements this algorithm.

Keywords: anomaly detection; topological data analysis; persistent homology; extreme value theory; peak over thresholds; generalized Pareto distribution (search for similar items in EconPapers)
JEL-codes: C55 C65 C87 (search for similar items in EconPapers)
Pages: 26
Date: 2021
New Economics Papers: this item is included in nep-ore
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.monash.edu/business/ebs/research/publications/ebs/wp02-2021.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:msh:ebswps:2021-2

Ordering information: This working paper can be ordered from
http://business.mona ... -business-statistics

Access Statistics for this paper

More papers in Monash Econometrics and Business Statistics Working Papers from Monash University, Department of Econometrics and Business Statistics PO Box 11E, Monash University, Victoria 3800, Australia. Contact information at EDIRC.
Bibliographic data for series maintained by Professor Xibin Zhang ().