Home | Publications | KS19

Insights Into a Running Clockwork: On Interactive Process-Aware Clustering

MCML Authors

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

In recent years the demand for having algorithms which provide not only their results, but also add explainability up to a certain extent increased. In this paper we envision a class of clustering algorithms where the users can interact not only with the input or output but also intercept within the very clustering process itself, which we coin with the term process-aware clustering. Further we aspire to sketch the challenges emerging with such type of algorithms, such as the need of adequate measures which evaluate the progression through the computation process of a clustering method. Beyond the explainability on how the results are generated, we propose methods tailored at systematically analyzing the hyperparameter space of an algorithm, determining in a more ordered fashion suitable hyperparameters rather then applying a trial-and-error schema.

inproceedings