Classification, Predicting Categories
A spam filter never answers 0.72; it answers spam or not-spam, a name from a short list. The same filter also knows how sure it is, and hands you a dial for which wrong picks you can live with.
- ▸The filter never says 'almost spam.' It says spam or not, and a wrong pick costs full price.
- ▸Two emails both flagged spam can be 55% and 95% sure; the label alone hides which.
- ▸No cut-off catches every spam and spares every real email. You choose which mistake to make.
BUILDS ON
An email lands and the filter has to commit to one answer: spam, or not. In the last topic the answer was a number you could be a little off on, guess $410k for a $400k house and you were basically right. Here there is no basically. The email is spam or it is not, and dropping a real email into the spam pile is not a near-miss, it is simply the wrong answer. The output is a name pulled from a fixed, short list, never a value sitting somewhere in between.
DROP EACH EMAIL INTO A BIN, THERE IS NO THIRD SLOT