How can I optimize the Naive Bayes classifier in Weka?

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

How can I optimize the Naive Bayes classifier in Weka?

eman.a.y

 

 

In WEKA to optimize the classifiers to get the best accuracy, we tune the parameters.

for example, I tuned 2 parameters FOR J48 (C and M) and I got good accuracy.

 

 

About Naive Bayes, when I asked someone he said there is no parameter for Naive Bayes classifier that can we optimize.

My questions are:

How can I optimize the Naive Bayes classifier in Weka?

Is Naive Bayes have parameters?

 

please, I need illustration in these two questions.

Thank you

Sent from Mail for Windows 10

 


_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: How can I optimize the Naive Bayes classifier in Weka?

Eibe Frank-3
NaiveBayes only has three possible configurations that affect the model (assuming you have numeric attributes in your data): assuming a class-conditional normal density (the default), using a kernel density estimate instead, or using discretisation. I would perhaps just specify the three different versions of NaiveBayes as base classifiers in MultiScheme, which will pick the best one using internal cross-validation.

Cheers,
Eibe



On Fri, Jul 26, 2019 at 8:56 AM Eman Abdalraheem <[hidden email]> wrote:

 

 

In WEKA to optimize the classifiers to get the best accuracy, we tune the parameters.

for example, I tuned 2 parameters FOR J48 (C and M) and I got good accuracy.

 

 

About Naive Bayes, when I asked someone he said there is no parameter for Naive Bayes classifier that can we optimize.

My questions are:

How can I optimize the Naive Bayes classifier in Weka?

Is Naive Bayes have parameters?

 

please, I need illustration in these two questions.

Thank you

Sent from Mail for Windows 10

 

_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.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: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: How can I optimize the Naive Bayes classifier in Weka?

eman.a.y
Thanks, Eibe for the clarification
so this is what I did as shown I change the" use kernel density estimate "
to true then I have good accuracy than the last one :) is that mean I did
optimization?



Untitled2.jpg <http://weka.8497.n7.nabble.com/file/t6877/Untitled2.jpg>  

and if you can please I need more clarification about the previous way for
optimizing.
because I did not understand about what you say " I would perhaps just
specify the three different versions of NaiveBayes as base classifiers in
MultiScheme, which will pick the best one using internal cross-validation."

Thanks



--
Sent from: http://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: How can I optimize the Naive Bayes classifier in Weka?

Eibe Frank-3
To avoid optimistic bias in your performance estimates due to parameter tuning on the test data (which is probably what you are doing), use something like MultiScheme instead. It will use internal cross-validation. Please take the MOOCs if you need further explanation of why it is problematic to optimise parameters based on the data that is also used to evaluate performance.

Cheers,
Eibe

On Fri, Jul 26, 2019 at 3:03 PM eman.a.y <[hidden email]> wrote:
Thanks, Eibe for the clarification
so this is what I did as shown I change the" use kernel density estimate "
to true then I have good accuracy than the last one :) is that mean I did
optimization?



Untitled2.jpg <http://weka.8497.n7.nabble.com/file/t6877/Untitled2.jpg

and if you can please I need more clarification about the previous way for
optimizing.
because I did not understand about what you say " I would perhaps just
specify the three different versions of NaiveBayes as base classifiers in
MultiScheme, which will pick the best one using internal cross-validation."

Thanks



--
Sent from: http://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.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: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html