Applying SMOTE

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

Applying SMOTE

andria lan
Hi all,

Is this method of using SMOTE acceptable for a regression problems?

Thanks in advance. 

Andria 


_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: 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: Applying SMOTE

Eibe Frank-2
Administrator
You could try it…

One way would be to use RegressionByDiscretization -> MultiClassClassifier -> FilteredClassifier -> SMOTE

There is also some work on a version of SMOTE for regression:

https://researchcommons.waikato.ac.nz/handle/10289/8518

However, AFAIK, this method not currently available in WEKA.

Cheers,
Eibe

> On 26/03/2018, at 1:32 PM, Andria Lan <[hidden email]> wrote:
>
> Hi all,
>
> Is this method of using SMOTE acceptable for a regression problems?
>
> Thanks in advance.
>
> Andria
>
> _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> List info and subscription status: 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]
List info and subscription status: 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: Applying SMOTE

andria lan
Thank you so much for the fast reply, Eibe. 

One way would be to use RegressionByDiscretization -> MultiClassClassifier -> FilteredClassifier -> SMOTE

I have tried this way with the "cpu" dataset and had this error: Cannot use 0 neighbors!

Note that I used the "Logistic" classifier as a base calssifier of "FilteredClassifier" (SMOTE is the filter of FilteredClassifier as well).

Why I had this error, and how to solve it?

Thanks once again. 

Andira 

There is also some work on a version of SMOTE for regression:

https://researchcommons.waikato.ac.nz/handle/10289/8518

However, AFAIK, this method not currently available in WEKA.

Cheers,
Eibe

> On 26/03/2018, at 1:32 PM, Andria Lan <[hidden email]> wrote:
>
> Hi all,
>
> Is this method of using SMOTE acceptable for a regression problems?
>
> Thanks in advance.
>
> Andria
>
> _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> List info and subscription status: 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]
List info and subscription status: 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]
List info and subscription status: 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: Applying SMOTE

Eibe Frank-3
My guess is that the equal-width discretization of the target attribute has created one or more intervals ("bins") without any actual values in it. RegressionByDiscretization has an option to "deleteEmptyBins" (-E). Try turning that on.

Cheers,
Eibe

On Wed, Mar 28, 2018 at 5:45 AM, Andria Lan <[hidden email]> wrote:
Thank you so much for the fast reply, Eibe. 

One way would be to use RegressionByDiscretization -> MultiClassClassifier -> FilteredClassifier -> SMOTE

I have tried this way with the "cpu" dataset and had this error: Cannot use 0 neighbors!

Note that I used the "Logistic" classifier as a base calssifier of "FilteredClassifier" (SMOTE is the filter of FilteredClassifier as well).

Why I had this error, and how to solve it?

Thanks once again. 

Andira 

There is also some work on a version of SMOTE for regression:

https://researchcommons.waikato.ac.nz/handle/10289/8518

However, AFAIK, this method not currently available in WEKA.

Cheers,
Eibe

> On 26/03/2018, at 1:32 PM, Andria Lan <[hidden email]> wrote:
>
> Hi all,
>
> Is this method of using SMOTE acceptable for a regression problems?
>
> Thanks in advance.
>
> Andria
>
> _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> List info and subscription status: 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]
List info and subscription status: 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]
List info and subscription status: 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]
List info and subscription status: 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: Applying SMOTE

Anatoliy
Hi Eibe!
Are there any changes in the past time when applying SMOTE to regression in
WEKA?
The process of using RegressionByDiscretization is not clear to me yet.
For example, when I use SMOTE by itself, I can visually control the width of
the bins and the amount of data in each bin. For example, if SMOTE refuses
to work, then this means that the number of neighbors exceeds the amount of
data in some bin. But when using RegressionByDiscretization I can't control
it, there are a lot of parameters.

regards
Anatoliy



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
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