You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@mahout.apache.org by Ramprakash Ramamoorthy <yo...@gmail.com> on 2011/12/30 14:19:17 UTC

Smoothing parameter in naive bayes classifier

Dear all,

     I see that I can set an alpha(smoothing parameter) to the Naive Bayes
classifier in apache mahout. The default value of alpha has been set to 1
and is an optional parameter. How does this alpha value affect the
classification results, can I just ignore it and use the default value or
should I change it for optimal classification?

-- 
With Thanks and Regards,
Ramprakash Ramamoorthy,
SASTRA University
+91 9626975420

Re: Smoothing parameter in naive bayes classifier

Posted by Lance Norskog <go...@gmail.com>.

There are two Bayes classifiers. 'TrainClassifier' and
'TestClassifier' are the "old" implementation. These apps take an
alpha parameter, but I cannot find where they consult it. Changing it
has no effect on the algorithm's performance in the 20 Newsgroups
example.

The ASF Emails example script examples/bin/asf-email-examples.sh uses
the new Bayes classifier, and changing Alpha does have an effect.

On Fri, Dec 30, 2011 at 10:13 AM, Ted Dunning <te...@gmail.com> wrote:
> This paper should tell you how alpha works:
> http://qwone.com/~jason/papers/sm-thesis.pdf
>
> On Fri, Dec 30, 2011 at 5:19 AM, Ramprakash Ramamoorthy <
> youngestachiever@gmail.com> wrote:
>
>> Dear all,
>>
>>     I see that I can set an alpha(smoothing parameter) to the Naive Bayes
>> classifier in apache mahout. The default value of alpha has been set to 1
>> and is an optional parameter. How does this alpha value affect the
>> classification results, can I just ignore it and use the default value or
>> should I change it for optimal classification?
>>
>> --
>> With Thanks and Regards,
>> Ramprakash Ramamoorthy,
>> SASTRA University
>> +91 9626975420
>>

-- 
Lance Norskog
goksron@gmail.com

Re: Smoothing parameter in naive bayes classifier

Posted by Ted Dunning <te...@gmail.com>.

This paper should tell you how alpha works:
http://qwone.com/~jason/papers/sm-thesis.pdf

On Fri, Dec 30, 2011 at 5:19 AM, Ramprakash Ramamoorthy <
youngestachiever@gmail.com> wrote:

> Dear all,
>
>     I see that I can set an alpha(smoothing parameter) to the Naive Bayes
> classifier in apache mahout. The default value of alpha has been set to 1
> and is an optional parameter. How does this alpha value affect the
> classification results, can I just ignore it and use the default value or
> should I change it for optimal classification?
>
> --
> With Thanks and Regards,
> Ramprakash Ramamoorthy,
> SASTRA University
> +91 9626975420
>