C-Command Software Forum

Do I Need To Retrain?

After resetting my History today (Michael recommended this within minutes after a crash log was sent), I noticed that my corpus is 5,057:17,973 ham:spam. That’s 78% spam, with what seems to be an inordinate number of messages.

Since the reset occurred I don’t know what the accuracy was, but I’ve not really noticed a problem. Should I still reset the corpus and retrain?

Don

You can check the recent accuracy by clicking the “Set Date…” button in the Statistics window. There are basically two reasons to reset the corpus:

  1. If the accuracy is unsatisfactory.
  2. If the launch time or processing speed is slow and
    the corpus has more than 5,000 or so messages.

If you’re happy with the way it’s working, there’s no need to do anything.