Davies-Bouldin Index

INFO

Calculates the ratio between intra-cluster and inter-cluster distances. A lower index value indicates better clustering.

How It Works

$$DB=\frac{1}{k}\sum _{i=1}^k\underset{j\ne i}{\max }\left(\frac{S_i+S_j}{M_{ij}}\right)$$

• $S_i$: Average distance within cluster $i$
• $M_{ij}$: Distance between centroids of clusters $i$ and $j$
• $k$: Number of clusters

What to Look For

• Lower DBI = better clustering
• Sensitive to both compactness and separation
• Useful for comparing clustering algorithms

Application Models

• K-Means Clustering
• Hierarchical Clustering
• Density-Based Spatial Clustering of Application with Noise (DBSCAN)
• Gaussian Mixture Models