Impact setting role to

Carlo

Hi Community,

I am asking myself if I set for one attribute the "label"-role when clustering. The only difference I can see is that the attribute does not appear anymore in the results sheets (centroid table, plot).

How does rapidminer takes the "label"-role into account when k-means clustering is performed?

Thanks for your explanations

Best regards,
Carlo

MartinLiebig

Hi @Carlo ,

since you do not need a label for clustering we will handle the label role like any other special role - it is ignored.

Best,

Martin

Papad

Hi,
I'm not sure if set role operator is needed when you select Clusters. I used set role for an attribute but when I selected clusters it disappeared. Instead, when I select predict, it gives you the option to select what to you want to predict so you choose a "label".
Furthermore when you want clusters, your dataset is will be splitted based on some criteria generally(some general correlations ) and not on a label.
I don't know if it helps, hope it does. Although I am not sure about my opinion so get others opinions too!

MartinLiebig

Hi @Carlo ,

since you do not need a label for clustering we will handle the label role like any other special role - it is ignored.

Best,

Martin

Papad

Hi,
I'm not sure if set role operator is needed when you select Clusters. I used set role for an attribute but when I selected clusters it disappeared. Instead, when I select predict, it gives you the option to select what to you want to predict so you choose a "label".
Furthermore when you want clusters, your dataset is will be splitted based on some criteria generally(some general correlations ) and not on a label.
I don't know if it helps, hope it does. Although I am not sure about my opinion so get others opinions too!