does anyone have an idea about how to evaluate multi-label classifiers? ie: when your label is not a single class but it can be multiple ones i have tried calculating the Jaccard index for each observation then summing all the values but i don't think that s a good metric.
$\begingroup$
$\endgroup$
1
-
$\begingroup$ A confusion matrix is a typical tool to understand the performance of multiclass classifiers. $\endgroup$– noeMay 11 at 15:16
Add a comment
|