Dimensionality-Aware Outlier Detection: Theoretical and Experimental Analysis

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Anderberg, Alastair, Bailey, James, Campello, Ricardo J. G. B., Houle, Michael E., Marques, Henrique O., Radovanović, Miloš, Zimek, Arthur
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913322893312000
author Anderberg, Alastair
Bailey, James
Campello, Ricardo J. G. B.
Houle, Michael E.
Marques, Henrique O.
Radovanović, Miloš
Zimek, Arthur
author_facet Anderberg, Alastair
Bailey, James
Campello, Ricardo J. G. B.
Houle, Michael E.
Marques, Henrique O.
Radovanović, Miloš
Zimek, Arthur
contents We present a nonparametric method for outlier detection that takes full account of local variations in intrinsic dimensionality within the dataset. Using the theory of Local Intrinsic Dimensionality (LID), our 'dimensionality-aware' outlier detection method, DAO, is derived as an estimator of an asymptotic local expected density ratio involving the query point and a close neighbor drawn at random. The dimensionality-aware behavior of DAO is due to its use of local estimation of LID values in a theoretically-justified way. Through comprehensive experimentation on more than 800 synthetic and real datasets, we show that DAO significantly outperforms three popular and important benchmark outlier detection methods: Local Outlier Factor (LOF), Simplified LOF, and kNN.
format Preprint
id arxiv_https___arxiv_org_abs_2401_05453
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Dimensionality-Aware Outlier Detection: Theoretical and Experimental Analysis
Anderberg, Alastair
Bailey, James
Campello, Ricardo J. G. B.
Houle, Michael E.
Marques, Henrique O.
Radovanović, Miloš
Zimek, Arthur
Machine Learning
Artificial Intelligence
68T99 (Primary) 62G07, 62G32, 62H30 (Secondary)
We present a nonparametric method for outlier detection that takes full account of local variations in intrinsic dimensionality within the dataset. Using the theory of Local Intrinsic Dimensionality (LID), our 'dimensionality-aware' outlier detection method, DAO, is derived as an estimator of an asymptotic local expected density ratio involving the query point and a close neighbor drawn at random. The dimensionality-aware behavior of DAO is due to its use of local estimation of LID values in a theoretically-justified way. Through comprehensive experimentation on more than 800 synthetic and real datasets, we show that DAO significantly outperforms three popular and important benchmark outlier detection methods: Local Outlier Factor (LOF), Simplified LOF, and kNN.
title Dimensionality-Aware Outlier Detection: Theoretical and Experimental Analysis
topic Machine Learning
Artificial Intelligence
68T99 (Primary) 62G07, 62G32, 62H30 (Secondary)
url https://arxiv.org/abs/2401.05453