Calculator Inputs
Use the responsive input grid below. Large screens show three columns, medium screens show two, and phones show one.
Example Data Table
This example demonstrates three clusters using diameters and pairwise minimum distances.
| Cluster | Size | Diameter |
|---|---|---|
| Cluster A | 20 | 2.30 |
| Cluster B | 18 | 1.80 |
| Cluster C | 22 | 2.10 |
| Pair | Minimum Distance |
|---|---|
| Cluster A to Cluster B | 4.50 |
| Cluster A to Cluster C | 5.20 |
| Cluster B to Cluster C | 4.10 |
Example Dunn Index: 4.10 ÷ 2.30 = 1.7826
Formula Used
Dunn Index = Minimum Inter-Cluster Distance ÷ Maximum Intra-Cluster Diameter
The numerator captures the smallest separation between any two clusters. The denominator captures the largest within-cluster diameter among all clusters.
A larger result usually indicates a better clustering structure because clusters are farther apart relative to their internal spread.
How to Use This Calculator
- Select how many clusters you want to evaluate.
- Enter a label, diameter, and optional size for each cluster.
- Provide the minimum distance for every cluster pair.
- Click the calculate button to display the result above the form.
- Review the interpretation, summary table, and export options.
Frequently Asked Questions
1. What does the Dunn Index measure?
It compares the smallest distance between clusters with the widest internal cluster diameter. Higher values suggest more compact and better-separated clusters.
2. Is a higher Dunn Index always better?
Usually yes, because a larger score means stronger separation relative to spread. Still, compare results with the same distance definition and clustering method.
3. What if one cluster is very wide?
A single wide cluster raises the denominator and can lower the whole score. That often signals poor compactness or outlier influence.
4. Do I need raw coordinates to use this page?
No. This version works from prepared summary statistics: cluster diameters and pairwise minimum inter-cluster distances.
5. Can I compare different clustering models?
Yes. Enter each model's cluster summary values separately, then compare the resulting Dunn Index values under the same distance basis.
6. Why are optional cluster sizes included?
Sizes do not change the standard Dunn formula here. They help document the clustering setup in exports and reports.
7. What distance values should I enter?
Use the minimum separation between each cluster pair and the maximum diameter inside each cluster, calculated from your chosen distance metric.
8. Can this score replace all validation checks?
No. It is useful, but pairing it with silhouette or Davies-Bouldin analysis gives a broader view of clustering quality.