Instruction: Describe metrics and methods used to assess the quality of clusters formed by a clustering algorithm.
Context: This question tests the candidate's knowledge on evaluating unsupervised learning models, particularly how to measure the effectiveness of clustering when true labels are not known.
Official answer available
Preview the opening of the answer, then unlock the full walkthrough.
The way I'd approach it in an interview is this: I evaluate clustering in two layers. First, I use internal measures like cohesion, separation, silhouette score, or stability across runs to understand whether the clusters are structurally meaningful. If I have...