The confusion matrix tells the truth faster — step 6 of 7
Write confusion_matrix(cases). Each case is a dict with predicted
and actual keys (values are "fraud" or "legit"). Return a dict
with the four counts in this order: tp, fp, tn, fn, where
"fraud" is the positive class.
- tp: predicted fraud, actual fraud
- fp: predicted fraud, actual legit
- tn: predicted legit, actual legit
- fn: predicted legit, actual fraud