MORe++: K-Means Based Outlier Removal on High-Dimensional Data

MCML Authors

Anna Beer

Dr.

* Former Member

→ Group Thomas Seidl
 Database Systems, Data Mining and AI

Thomas Seidl

Prof. Dr.

Director

Database Systems, Data Mining and AI

Abstract

MORe++ is a k-Means based Outlier Removal method working on high dimensional data. It is simple, efficient and scalable. The core idea is to find local outliers by examining the points of different k-Means clusters separately. Like that, one-dimensional projections of the data become meaningful and allow to find one-dimensional outliers easily, which else would be hidden by points of other clusters. MORe++ does not need any additional input parameters than the number of clusters k used for k-Means, and delivers an intuitively accessible degree of outlierness. In extensive experiments it performed well compared to k-Means-- and ORC.

inproceedings BLS19

---

SISAP 2019

12th International Conference on Similarity Search and Applications. Newark, New York, USA, Oct 02-04, 2019.

---

Authors

A. Beer • J. Lauterbach • T. Seidl

---

Links

DOI

---

Research Area

 A3 | Computational Models