A novel CVI: Distance-based Separability Index (DSI)

This is an example of computing Cluster Validity Index (CVI) on datasets after clustering.

It includes codes to compute the distance-based separability index (DSI), which is used in another study: "Data Separability for Neural Network Classifiers and the Development of a Separability Index" (Preprint).

Related paper

An Internal Cluster Validity Index Using a Distance-based Separability Measure

Citation S. Guan and M. Loew, "An Internal Cluster Validity Index Using a Distance-based Separability Measure," 2020 IEEE 32nd International Conference on Tools with Artificial Intelligence (ICTAI), Baltimore, MD, USA, 2020, pp. 827-834, doi: 10.1109/ICTAI50040.2020.00131.

Paper Arxiv Video

Data Separability for Neural Network Classifiers and the Development of a Separability Index

Arxiv

Results

DSI scores of the handwritten digits dataset after applying several clustering methods

Output in results.txt.

• 0.4933647148073444 ture labels
• 0.4836361123766773 K-Means
• 0.5716958115555282 Spectral Clustering
• 0.5471394199023191 Birch
• 0.4697710019443232 Gaussian Mixture (EM)