error on determine silhouette for k-means clustering

livnhn

Hello

I'm using rapidminer version 8.1

I want to calculate the silhouette value for k-means clustering

For this purpose
I downloaded the plugin from the following URL:
http://korek.name/web/moje-tvorba/rapidminer-clustering_performance_plugin-average_silhouette-cophenetic_coefficient

 

My process image is as follows:

 

In this process, the following error occurs:

 

 

Does anyone know where the problem is?

 

thanks
regards

David_A

Hi,

 

I don't know that specific extension.

But from the error code my guess would be, that it might require a numerical cluster attribute. What you could try is to add a Nominal to Numerical  Operator after the clustered ExampleSet and change the cluster Attribute to unique integers (remember to check "include special attributes").

No guarantee but it might do the trick.

 

Best,
David

 

sgenzer

hi @livnhn- that looks like a VERY old extension. I would not find it surprising if it produced errors with RapidMiner 8. If you're just trying to do performance calculations with k-means clustering, there are several native operators that can help:

 

 

Scott

 

Muhammed_Fatih_

Hi @sgenzer, 

which one of the natives ones would you recommend to evaluate the k for k-means? 

Cheers!

lionelderkrikor

Hi Muhammed_Fatih_,

You mean determine a priori the best number of clusters k ?
You can use the Performance (Cluster Distance Performance) operator and set the main criterion as Average within centroid distance in the parameters.
Then you can use an Optimization loop to plot the Average within centroid distance according to k (the number of clusters).
The method is explained in this thread : 

https://community.rapidminer.com/discussion/comment/61654#Comment_61654

Hope this helps,

Regards,

Lionel 

Muhammed_Fatih_

Hi @lionelderkrikor, 

yes, the a priori determination of clusters k. Thank you for the hint!