RE: [Classifier4j-devel] Next Step after Training?

[Nick L. <nl...@es...>]

> 
> 
> Right - your first two for-loops are training your
> classifier to learn which messages are your
> whitelist messages.  Once this is done you can
> test any other message against your training
> to get a "rating".  When you run the classify()
> method it will return a value between 0.0 and 1.0.
> 
> 0 meaning that the new message you ran classify
> on is definitely a blacklist message... 1.0 meaning
> its definitely a whitelist message.
> 
> At least this is the way I'm using Classifier4J..
> not sure if its the absolute correct way :)
> 
> My rules are similar to SpamBayes in that
> I mark anything with a 0.9 and above as a definite
> match.. anything below that is considered a partial
> match (I use C4J as a spam filter against email msgs).
> 
> Nick - correct me if Im wrong?  I'm no C4J expert,
> but using it this way seems to work pretty well
> for me.
> 

Yes, that is exactly correct.

You can use the IClassifier.isMatch(String) to do the same thing - each
instance of the classifier has a setCutoff() method (I think that's the
name) to set the exact point above which anything will be marked as spam.