Note that the multilabel and multiclass classifications sound similar, but they are two different things.
All multilabel MultilabelMetrics() method is trying to accomplish is to map a number of inputs (x) to a binary vector (y) rather than numerical values in a typical classification system.
The important metrics associated with the multilabel classification are (see the preceding code):
- Accuracy
- Hamming loss
- Precision
- Recall
- F1
A full explanation of each parameter is out of scope, but the following link provides a short treatment for the multilabel metrics: