mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From gaurav garg <gaurav.gar...@gmail.com>
Subject Re: Nave Bayes Classifier and probability calculation for a single test example
Date Wed, 04 May 2011 13:29:04 GMT
Svetlomir,

Unfortunately, naive bayes does not output a probability score. Ted Dunning
has explained the bayes score in a separate email thread  with subject
"Scoring issue".  I will also forward you that email chain.

I am copying here the excerpts:

************************************************
Naive Bayes does not produce a viable probability score (SGD tries to, in
contrast).

These scores are comparable between categories for a single document.
 Comparing scores for different documents is not easy.

You may be able to calibrate the scores, but I don't know if that will work.

Have you read Renny's paper?

http://citeseer.ist.psu.edu/viewdoc/summary?doi=10.1.1.13.8572

************************************************




Thanks
gaurav

On Wed, May 4, 2011 at 6:16 PM, Svetlomir Kasabov <
skasab2s@smail.inf.fh-brs.de> wrote:

> Hello Gaurav,
>
> thanks for your reply. I tried your hint but it still does not lead me to a
> probability for a single training example. How can I use the score
> of a classified training example to get a probability ?
>
> Thanks a lot and many regards.
>
> Svetlomir.
>
>
>  Sent from BlackBerry® on Airtel
>>
>> -----Original Message-----
>> From: gaurav garg<gaurav.garg28@gmail.com>
>> Date: Wed, 4 May 2011 00:37:53
>> To:<user@mahout.apache.org>
>> Reply-To: user@mahout.apache.org
>> Cc:<mahout-user@lucene.apache.org>
>> Subject: Re: Nave Bayes Classifier and probability calculation for a
>> single
>>  test example
>>
>> Hi Svetlomir,
>>
>> You can use ClassifierContext class (
>>
>> https://builds.apache.org/hudson/job/Mahout-Quality/javadoc/index.html?org/apache/mahout/common/StringTuple.html
>> )
>> to get the top N matching result and their respective scores.
>>
>> Hope it helps.
>>
>> Thanks
>> Gaurav
>>
>
>
>


-- 
Gaurav Garg

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message