Comments on: Bayes Classification in Ruby

By: Mike Subelsky

Mike Subelsky — Fri, 06 Jun 2008 15:34:57 +0000

I was able to get this working; all you have to do is run “Marshal.dump(classifier)” and store the output to a text column in your database. For MySQL I got errors unless I made the column type “longtext”, a la:

change_column :tags, :bayes_data, :longtext

Then to restore the classifier, you just do:

Marshal.load(bayes_data) and voila!

By: Ilya Grigorik

Ilya Grigorik — Fri, 06 Jun 2008 12:10:16 +0000

Thanks for the tip Mike!

By: Mike Subelsky

Mike Subelsky — Fri, 30 May 2008 13:47:10 +0000

…also dump total words, forgot that one as well.

By: Mike Subelsky

Mike Subelsky — Fri, 30 May 2008 13:00:48 +0000

For the people who want to persist to a database, you can’t use SnapshotMadeline. You need to add custom dump and load methods to the Classifier class. Looking at the implementation of the Bayes classifier, it seems like only the @categories instance variable needs to be dumped and reloaded to reconstitute the learning set. This fragment from the Pickaxe should help you get started: Marshalling in Ruby

By: Ilya Grigorik

Ilya Grigorik — Sat, 03 May 2008 23:36:39 +0000

Daniel, you might want to contact the author with your implementation then — that way everyone gets the updated gem.

By: Daniel

Daniel — Mon, 28 Apr 2008 15:47:24 +0000

as I already wrote in this thread I have reviewed the source line by line; the awful results just ringed a bell;

sorry but this program implements anything but a bayesian filter.

By: Ilya Grigorik

Ilya Grigorik — Tue, 22 Apr 2008 03:53:16 +0000

:) just because the algorithm did not give good accuracy on a dataset, doesn’t mean the algorithm is broken. Having said that, if the algorithm is not working properly, and you can contribute to the source (sourceforge) to make it better, that would be awesome!

By: Daniel

Daniel — Thu, 17 Apr 2008 21:28:01 +0000

that’s what I imagined: a blind “implementation” of an algorithm the programmer does not seem to understand.

By: Ilya Grigorik

Ilya Grigorik — Wed, 16 Apr 2008 06:59:47 +0000

Hummm. Daniel, it’s been a while since I’ve touched that code, but the implementation was directly from the source (the ‘AI bible’, 2nd edition).

By: Daniel

Daniel — Mon, 07 Apr 2008 13:37:10 +0000

“Both gems implement the direct algorithms - probabilities and all”

Then please explain the presence of the logarithmic function applied in the “classification” function… Or maybe the “negative probabilities” it outputs.