The Davies-Bouldin index (BDI) is the ratio of intra-cluster distance to inter-cluster distance. A lower DBI value means better clustering results. This can be calculated as follows:
Here, the following applies:
- n: The number of clusters
- ci: The centroid of cluster i
- σi: The intra-cluster distance or average distance of all cluster items from the centroid ci
- d(ci, cj): The inter-cluster distance between two cluster centroids ci and cj