Hands-On Unsupervised Learning with Python

Cophenetic correlation as a performance metric

Hierarchical clustering performance can be evaluated by using any of the methods presented in the previous chapters. However, in this particular case, a specific measure (that doesn't require the ground truth) can be employed. Given a proximity matrix, P, and a linkage, L, a couple of samples, x_i and x_j ∈ X, are always assigned to the same cluster at a certain hierarchical level. Of course, it's important to remember that in the agglomerative scenario, we start with n different clusters and we end up with a single cluster equivalent to X. Moreover, as two merged clusters become a single one, two samples belonging to a cluster will always continue to belong to the same enlarged cluster until the ...