Machine Learning Algorithms

Calinski-Harabasz index

Another method that is based on the concept of dense and well-separated clusters is the Calinski-Harabasz index. To build it, we need first to define the inter cluster dispersion. If we have k clusters with their relative centroids and the global centroid, the inter-cluster dispersion (BCD) is defined as:

In the above expression, n_k is the number of elements belonging to the cluster k, mu (the Greek letter in the formula) is the global centroid, and mu_i is the centroid of cluster i. The intracluster dispersion (WCD) is defined as:

The Calinski-Harabasz index is defined as the ratio between BCD(k) and WCD(k):

As we ...