Holistic Assessment of Structure Discovery Capabilities of Clustering Algorithms
Abstract
Existing cluster validity indices often possess a similar bias as the clustering algorithm they were introduced for, e.g. to determine the optimal number of clusters. We suggest an efficient and holistic assessment of the structure discovery capabilities of clustering algorithms based on three criteria. We determine the robustness or stability of cluster assignments and interpret it as the confidence of the clustering algorithm in its result. This information is then used to label the data and evaluate the consistency of the stability-assessment with the notion of a cluster as an area of dense and separated data. The resulting criteria of stability, structure and consistency provide interpretable means to judge the capabilities of clustering algorithms without the typical biases of prominent indices, including the judgment of a clustering tendency.
Cite
Text
Höppner and Jahnke. "Holistic Assessment of Structure Discovery Capabilities of Clustering Algorithms." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2019. doi:10.1007/978-3-030-46150-8_14Markdown
[Höppner and Jahnke. "Holistic Assessment of Structure Discovery Capabilities of Clustering Algorithms." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2019.](https://mlanthology.org/ecmlpkdd/2019/hoppner2019ecmlpkdd-holistic/) doi:10.1007/978-3-030-46150-8_14BibTeX
@inproceedings{hoppner2019ecmlpkdd-holistic,
title = {{Holistic Assessment of Structure Discovery Capabilities of Clustering Algorithms}},
author = {Höppner, Frank and Jahnke, Maximilian},
booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases},
year = {2019},
pages = {223-239},
doi = {10.1007/978-3-030-46150-8_14},
url = {https://mlanthology.org/ecmlpkdd/2019/hoppner2019ecmlpkdd-holistic/}
}