Machine Learning With Go

External clustering evaluation

If we have a ground truth or gold standard for our clusters, then we can utilize a variety of external clustering evaluation techniques. This ground truth or gold standard means that we have access to, or can get (via manual human annotation), a set of data points where the true or desired cluster labels have been annotated.

Often, we do not have access to this sort of clustering gold standard, and, thus, we will not cover these sorts of evaluation techniques in detail here. However, if interested or relevant, you can look into the Adjusted Rand index, Mutual Information, Fowlkes-Mallows scores, completeness, and V-measures, which are all relevant external clustering evaluation metrics (read more details about ...