[Spambayes] 2 Questions

[Meyer, Tony]

> if just trained my spambayes, will it score wrong if i have 
> 418 ham and only 189 spam?

Well, technically spambayes can't (except for bugs) score *wrong* - but the score might not match what you want.

It's really impossible for us to answer this.  How well spambayes works depends on many factors, not just the number of ham and spam that have been trained.  Other factors include how different (token-wise) your ham and spam are, how different your training data and classification data are, and many other things.

Really, the best thing would be to try it out and see how things go.  It'll certainly be better than blind chance, or no filtering at all. 

> (Maybe i should sign up for more 
> spam-crap so i get an even number to train?) 

Well, presumably you regularly get more, or there isn't much point in using spambayes.  Just train on it as it arrives, or train on anything misclassified or unsure.  You could also try reducing the amount of ham to match the spam more closely - more isn't always better.  In the end, though, you're the only one that can check how well your training data works.

=Tony Meyer