Weighted Clustering

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

Weighted Clustering

Meowgi
Hello,

Is there a way to add weights to attributes and cluster accordingly?

Thanks.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Weighted Clustering

Eibe Frank-2
Administrator
Only NaiveBayes in WEKA currently makes use of attribute weights. However, you can rescale the attributes according to their importance and this will affect the result of distance-based clustering algorithms. Just make sure you turn normalisation off when you do this, so the attributes are not put on the same scale again! There is a parameter in the distance functions in WEKA that allows you to turn normalisation off.

Cheers,
Eibe

> On 11/05/2017, at 2:56 AM, Meowgi <[hidden email]> wrote:
>
> Hello,
>
> Is there a way to add weights to attributes and cluster accordingly?
>
> Thanks.
>
>
>
> --
> View this message in context: http://weka.8497.n7.nabble.com/Weighted-Clustering-tp40573.html
> Sent from the WEKA mailing list archive at Nabble.com.
> _______________________________________________
> 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
|  
Report Content as Inappropriate

Re: Weighted Clustering

Meowgi
Thank you for answering.

How do i re scale the attributes appropriately? Is there a filter where i can do that? Like for example lets say i have 3 attributes:

Att1 - 50%
Att2 - 30%
Att3 - 20%

Is this possible? If so where can i find it?

Thanks
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Weighted Clustering

Eibe Frank-2
Administrator
You can use the MathExpression filter to scale an attribute.

The exact scaling function to apply depends on the distance function. It’s most straightforward in the case of Manhattan distance: if you have three attributes with weights 0.5, 0.3, and 0.2, just multiply each of the attributes’ values by the corresponding weight. In the case of the Euclidean distance, you could use the square root of the weight as the multiplier.

Cheers,
Eibe

> On 12/05/2017, at 12:37 AM, Meowgi <[hidden email]> wrote:
>
> Thank you for answering.
>
> How do i re scale the attributes appropriately? Is there a filter where i
> can do that? Like for example lets say i have 3 attributes:
>
> Att1 - 50%
> Att2 - 30%
> Att3 - 20%
>
> Is this possible? If so where can i find it?
>
> Thanks
>
>
>
>
> --
> View this message in context: http://weka.8497.n7.nabble.com/Weighted-Clustering-tp40573p40586.html
> Sent from the WEKA mailing list archive at Nabble.com.
> _______________________________________________
> 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
Loading...