you seem to think all classical ML models were supervised, which isn't true. and we have metrics for unsupervised approaches as well