Re: Feature Selection

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

Re: Feature Selection

Mark Hall

All of Weka’s attribute selection methods (with the exception of principal components, which is actually a feature transformer) are supervised. So, they cannot be used with the FilteredClusterer meta clusterer. If there is a class attribute in the data then you could first apply feature selection with respect to the class, and then apply clustering to the reduced data.

 

Cheers,

Mark.

 

On 9/01/19, 8:54 AM, "[hidden email] on behalf of Jovani T. de Souza" <[hidden email] on behalf of [hidden email]> wrote:

 

https://mailtrack.io/trace/mail/73c68d9e7b428525dea1d1011920a19321348f13.png?u=1864508

Hello friends.

 

Do you know if it is possible to work with feature selection where the correlation filter is integrating with clustering? 

 

Thank you very much !!

 

Mailtrack

Remetente notificado por
Mailtrack 08/01/19 17:50:37

 


_______________________________________________
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: Feature Selection

Jovani Souza
Mark,

My databases have class attribute. The steps would be as follows:
For example, if I chose to use CfsSubsetEval as the feature selection. I apply CfsSubsetEval (Supervisioned - Attribute Selection) and the generated result I apply the ADD Cluster - FilteredClusterer (Unsupervised). Is correct? 

Subsequently, can I check the hit rate of any algorithm?

Thanks again. 



Mailtrack Remetente notificado por
Mailtrack 15/01/19 22:35:05

Em ter, 15 de jan de 2019 às 20:51, Mark Hall <[hidden email]> escreveu:

All of Weka’s attribute selection methods (with the exception of principal components, which is actually a feature transformer) are supervised. So, they cannot be used with the FilteredClusterer meta clusterer. If there is a class attribute in the data then you could first apply feature selection with respect to the class, and then apply clustering to the reduced data.

 

Cheers,

Mark.

 

On 9/01/19, 8:54 AM, "[hidden email] on behalf of Jovani T. de Souza" <[hidden email] on behalf of [hidden email]> wrote:

 

https://mailtrack.io/trace/mail/73c68d9e7b428525dea1d1011920a19321348f13.png?u=1864508

Hello friends.

 

Do you know if it is possible to work with feature selection where the correlation filter is integrating with clustering? 

 

Thank you very much !!

 

Mailtrack

Remetente notificado por
Mailtrack 08/01/19 17:50:37

 

_______________________________________________
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