How does the Bayes classifier work?
The Bayes classifier runs on the IMAP servers just prior to delivery to a mailbox. After training with messages that you consider to be spam and ham (not-spam), the classifier will calculate the statistical probability that a message is ham, spam, or neither. A score can be assigned to each of those results and that score can be combined with the SpamAssassin score or the score used by itself to determine what to do with the message. The classifier is very effective in dealing with messages like ‘stock pump and dump’ spam that mutates rather quickly and does not contain the easily detected technical errors like most spam. The classifier is not enabled by default and it MUST be trained before use.