Class balancing with features selection

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

Class balancing with features selection

asadbtk
Hello

If we have to use class balancing and features selection, which should be performed first. I read two articles and one recommended class balancing first and another feature selection algorithms first. 

My second question is slightly different than this, if we use 10 fold cv and repeat it 10 times, can we say that we performed the experiments with 100 times repetition or it is just considered as 10 times? 

Best regards 

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

Re: Class balancing with features selection

Eibe Frank
Most likely, it depends on how you do attribute selection. For example, I would say that if you wanted to select attributes for a classifier to be trained on balanced data and apply the wrapper method for attribute selection, it would be best to also balance the data for attribute selection.

Regarding evaluation, it is best to state exactly what method was applied. In this case, you should state that 10-fold cross-validation was repeated 10 times to obtain 100 performance estimates (assuming you are using the default mode in the Experimenter and have not switched to the AveragingResultProducer in the advanced mode).

Cheers,
Eibe

On Fri, Feb 14, 2020 at 10:08 PM javed khan <[hidden email]> wrote:
Hello

If we have to use class balancing and features selection, which should be performed first. I read two articles and one recommended class balancing first and another feature selection algorithms first. 

My second question is slightly different than this, if we use 10 fold cv and repeat it 10 times, can we say that we performed the experiments with 100 times repetition or it is just considered as 10 times? 

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

Re: Class balancing with features selection

asadbtk
Thanks a lot Eibe for your useful information.

Best regards
Javed

On Mon, Feb 17, 2020 at 11:52 PM Eibe Frank <[hidden email]> wrote:
Most likely, it depends on how you do attribute selection. For example, I would say that if you wanted to select attributes for a classifier to be trained on balanced data and apply the wrapper method for attribute selection, it would be best to also balance the data for attribute selection.

Regarding evaluation, it is best to state exactly what method was applied. In this case, you should state that 10-fold cross-validation was repeated 10 times to obtain 100 performance estimates (assuming you are using the default mode in the Experimenter and have not switched to the AveragingResultProducer in the advanced mode).

Cheers,
Eibe

On Fri, Feb 14, 2020 at 10:08 PM javed khan <[hidden email]> wrote:
Hello

If we have to use class balancing and features selection, which should be performed first. I read two articles and one recommended class balancing first and another feature selection algorithms first. 

My second question is slightly different than this, if we use 10 fold cv and repeat it 10 times, can we say that we performed the experiments with 100 times repetition or it is just considered as 10 times? 

Best regards 
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to: To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit
https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
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
https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
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
https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html