Bayesian normality

From ASSPSMTP

Bayesian normality (also referred to as Bayesian norm) is a median value of the balance between ham and spam word-pairs in the Bayesian database. The most ideal ham to spam relationship for spam analysis is one-to-one, so the median Bayesian normality value is 1.0000.

The lower the normality value below 1.0000, the more ham there is than spam - and thus the greater the risk of false-negative spam matches. Inversely, the higher the value above 1.0000, the more spam there is than ham - and thus the greater the risk of false-positive ham matches.

Values between 0.5000 and 1.5000 are considered normal and healthy, although the ideal value is 1.0000. Anything outside of that range should be investigated as to why there is an offset, and should be remedied as soon as possible.

The Bayesian normality value is saved in the file rebuildrun.txt as part of the output from running the rebuildspamdb.pl script, which is ASSP's naive Bayes classifier.