Accepted paper at ICDM 2021 Workshop HDM 2021

Authors

Anna Beer, Lisa Stephan, Thomas Seidl

The 9th ICDM Workshop on High Dimensional Data Mining (HDM 2021)
in conjunction with the 21st IEEE International Conference on Data Mining (ICDM 2021),
07–10 December 2021, Auckland, New Zealand

Abstract

LUCKe allows any purely distance-based “classic” clustering algorithm to reliably find linear correlation clusters. An elaborated distance matrix based on the points' local PCA extracts all necessary information from high dimensional data to declare points of the same arbitrary dimensional linear correlation cluster as “similar”. For that, the points' eigensystems as well as only the relevant information about their position in space, are put together. LUCKe allows transferring known benefits from the large field of basic clustering to correlation clustering. Its applicability is shown in extensive experiments with simple representatives of diverse basic clustering approaches.