At the moment this function (and the cross-validation version) has an attribute called _minspread of type boolean. When True, the optimization selects, among the equally good-performing, the criterion that minimizes the difference between precision and recall; when False, it selects the minimum value of the criterion.
It could be interesting to change this attribute to a new attribute named prioritize:
min: it would select the minimum value of the criterion.
max: it would select the maximum value of the criterion.
_minspread: it would select the minimum difference between precision and recall.
At the moment this function (and the cross-validation version) has an attribute called _minspread of type boolean. When True, the optimization selects, among the equally good-performing, the criterion that minimizes the difference between precision and recall; when False, it selects the minimum value of the criterion.
It could be interesting to change this attribute to a new attribute named prioritize: