What is the limit to evaluate attributes?

classic Classic list List threaded Threaded
4 messages Options
JC
Reply | Threaded
Open this post in threaded view
|

What is the limit to evaluate attributes?

JC
Hi everyone!

I am looking for the limit to evaluate attributes with Java API for WEKA. I
am using Naive Bayes because is the only algorithm that let weight
attributes.

But *I am looking for the limit to weigth these attributes but I can't find
nothing*!
For example, I am using a double with the value 11 but I can put in the
value 0 or 1000.

Someone can help me with this?

Thanks in advance :)



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to: To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit %(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: What is the limit to evaluate attributes?

Eibe Frank-2
Administrator
WEKA does not impose a limit. However, when the weight is too large, you may run into numerical problems because of the limitations of double-precision floating point numbers.

The smallest positive double-precision value is Double.MIN_VALUE:


If p is the estimated probability of an observed attribute value given a particular classification, and w is the weight, then NaiveBayes will incorporate the attribute weight by calculating p^w and use that instead of p.

Cheers,
Eibe

On Fri, Sep 13, 2019 at 10:45 PM JC <[hidden email]> wrote:
Hi everyone!

I am looking for the limit to evaluate attributes with Java API for WEKA. I
am using Naive Bayes because is the only algorithm that let weight
attributes.

But *I am looking for the limit to weigth these attributes but I can't find
nothing*!
For example, I am using a double with the value 11 but I can put in the
value 0 or 1000.

Someone can help me with this?

Thanks in advance :)



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to: To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit %(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to: To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit %(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
JC
Reply | Threaded
Open this post in threaded view
|

Re: What is the limit to evaluate attributes?

JC
Thanks Eibe for your response.

Seeing this, the maximum value that can be given to an attribute is
1,7976931348623157E308 and the minimum value is 4.9E-324.

But I can't understand it well, this algorithm has to have an explanation to
give a weight to the attribute. If the weight is given to an attribute in a
project, for example 10 (double), this weight must have a reason, right?

Regards.



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to: To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit %(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: What is the limit to evaluate attributes?

Eibe Frank-2
Administrator
Giving an attribute an integer weight N, you effectively achieve the same effect as you would if you added N-1 copies to the dataset. For example, using a weight of 2 would be equivalent to introducing one copy of the attribute in the data.

Cheers,
Eibe

> On 17/09/2019, at 10:26 PM, JC <[hidden email]> wrote:
>
> Thanks Eibe for your response.
>
> Seeing this, the maximum value that can be given to an attribute is
> 1,7976931348623157E308 and the minimum value is 4.9E-324.
>
> But I can't understand it well, this algorithm has to have an explanation to
> give a weight to the attribute. If the weight is given to an attribute in a
> project, for example 10 (double), this weight must have a reason, right?
>
> Regards.
>
>
>
> --
> Sent from: https://weka.8497.n7.nabble.com/
> _______________________________________________
> Wekalist mailing list -- [hidden email]
> Send posts to: To unsubscribe send an email to [hidden email]
> To subscribe, unsubscribe, etc., visit %(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
> List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to: To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit %(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html