A classifier produces a probability that a given record is "positive" (here, that
is ).
The threshold (or "cut-off") is the probability above which the prediction is considered positive.
If set too low, it may predict too often, if set too high,
too rarely.
This chart shows how the overall performance metrics of the classifier vary depending on the threshold.
This model was evaluated with K-fold cross-test. The colored area around the curves show the double standard deviation of the curve, which indicates the stability of the model across folds. The narrower, the better.