Initially, you might want to train the spam filter manually to quickly build a database of spam and non-spam tokens, words, or short character sequences that are commonly found in spam or ham. To do this, you can manually forward messages as message/rfc822 attachments to the spam and non-spam mailboxes.
When zmtrainsa runs, these messages are used to teach the spam filter. Make sure you add a large enough sampling of messages to get accurate scores. To determine whether to mark messages as spam at least 200 known spams and 200 known hams must be identified.
