[prev in list] [next in list] [prev in thread] [next in thread] 

List:       wekalist
Subject:    Re: [Wekalist] NaiveBayesSimple.normalDens
From:       Eibe Frank <eibe () cs ! waikato ! ac ! nz>
Date:       2012-11-27 20:45:25
Message-ID: A35F6065-5C59-4AF2-B5E5-A981C49FF1F5 () cs ! waikato ! ac ! nz
[Download RAW message or body]

NaiveBayesSimple uses a normal density for numeric attributes. Probability density \
functions can have values > 1. They need to be nonnegative and integrate to 1 though.

Cheers,
Eibe

On 28 Nov 2012, at 09:16, Eduardo Correa Goncalves wrote:

> Hi,
> 
> I've been testing the class "NaiveBayesSimple" (Weka-3-6-8) against several \
> datasets. I have noticed that for some instances in a few of the datasets the \
> method "normalDens" returns a value above 1.0 for conditional probabilities (I \
> discovered that wrong probability values after debugging my code).  
> This error occurs, for example, in datasets such as "emotions.arff" and \
> "scene.arff", when I'm performing binary classification (below, 2 examples of \
> mistaken values returned by normalDens in the dataset "scene.arff").   
> 2.1060123249038063
> 1.7688895213725202
> 
> 
> 
> In spite of the fact that "NaiveBayesSimple" was created for educational purposes, \
> I would like to know why this kind of error occurs. Would it be due to underflow?  
> Cheers,
> 
> Eduardo
> 
> 
> 
> 
> _______________________________________________
> Wekalist mailing list
> Send posts to: Wekalist@list.scms.waikato.ac.nz
> List info and subscription status: \
> https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: \
> http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html



_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/=
listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.=
html


[prev in list] [next in list] [prev in thread] [next in thread]