Home  | Publications | LBH+21

AnyCORE - An Anytime Algorithm for Cluster Outlier REmoval

MCML Authors

Link to Profile Thomas Seidl PI Matchmaking

Thomas Seidl

Prof. Dr.

Director

Database Systems, Data Mining and AI

Abstract

We introduce AnyCORE (Anytime Cluster Outlier REmoval), an algorithm that enables users to detect and remove outliers at anytime. The algorithm is based on the idea of MORe++, an approach for outlier detection and removal that iteratively scores and removes 1d-cluster-outliers in n-dimensional data sets. In contrast to MORe++, AnyCORE provides continuous responses for its users and converges independent of cluster centers. This allows AnyCORE to perform outlier detection in combination with an arbitrary clustering method that is most suitable for a given data set. We conducted our AnyCORE experiments on synthetic and real-world data sets by benchmarking its variant with k-Means as the underlying clustering method versus the traditional batch algorithm version of MORe++. In extensive experiments we show that AnyCORE is able to compete with the related batch algorithm version.

inproceedings

LWDA 2021

Conference on Lernen. Wissen. Daten. Analysen. München, Germany, Sep 01-03, 2021.

Authors

A. Lohrer • A. Beer • M. Hünemörder • J. Lauterbach • T. Seidl • P. Kröger

Links

PDF

Research Area

 A3 | Computational Models

BibTeXKey: LBH+21

Back to Top