Grace - Limiting the Number of Grid Cells for Clustering High-Dimensional Data

MCML Authors

Anna Beer

Dr.

* Former Member

→ Group Thomas Seidl
 Database Systems, Data Mining and AI

Daniyal Kazempour

Dr.

* Former Member

→ Group Thomas Seidl
 Database Systems, Data Mining and AI

Thomas Seidl

Prof. Dr.

Director

Database Systems, Data Mining and AI

Abstract

Using grid-based clustering algorithms on high-dimensionaldata has the advantage of being able to summarize datapoints into cells, but usually produces an exponential number of grid cells. In this paper we introduce Grace (using textit{Gr}id which is textit{a}daptive for textit{c}lusttextit{e}ring), a clustering algorithm which limits the number of cells produced depending on the number of points in the dataset. A non-equidistant grid is constructed based on the distribution of points in one-dimensional projections of the data. A density threshold is automatically deduced from the data and used to detect dense cells, which are later combined to clusters. The adaptive grid structure makes an efficient but still accurate clustering of multidimensional data possible. Experiments with synthetic as well as real-world data sets of various size and dimensionality confirm these properties.

inproceedings

---

LWDA 2020

Conference on Lernen. Wissen. Daten. Analysen. Bonn, Germany, Sep 09-11, 2020.

---

Authors

A. Beer • D. Kazempour • J. Busch • A. Tekles • T. Seidl

---

Links

PDF

---

Research Area

 A3 | Computational Models