"Noob Problem: Clustering for selected Attributes"

uc4

Hello,
I am totally new to RM and have the following problem: I to cluster a number of tasks with respect to their runtime which is a given attribute. Unfortunately it is not clear to me how I can tell my k-Means operator that only the runtime attribute should be considered.
I know that I could use a "Select Attributes" before, but I do not want to delete the other attributes. So how can I realize this?
Where can I find examples?

With regards

Unknown

Hello,

You could use the "set role" operator to change the type of the attributes you want to ignore to "ignore1", "ignore2" and so on. You can type the text you want directly into the parameter and you have to use different role types for each attribute.

regards

Andrew

uc4

awchisholm wrote:
You could use the "set role" operator to change the type of the attributes you want to ignore to "ignore1", "ignore2" and so on. You can type the text you want directly into the parameter and you have to use different role types for each attribute.

Thank you very much for your answer.
I just could not find anything in the doc that says that only specific roles are used for clustering?

IngoRM

Hi,

I just could not find anything in the doc that says that only specific roles are used for clustering?

well, it does not state this explicitly for clustering (but for data transformations in general) but at the end of Section 3.2.1 is a statement that only regular attributes (without a special role) are used. There the manual states:

Tip: It is a basic rule of RapidMiner that operators from the group 'Data Transformation' are usually only executed on regular attributes, so on those without a special role. However, the operators offer an option called 'include special attributes' for this, meaning that the changes are also applied to those with a special role.

The same thing is basically true for all operators: the use the regular attributes or those with a specific role (like the "label" which is necessary for classification and regression learning). All others will be ignored.

Hope that helps and cheers,
Ingo

uc4

Thank you very much Ingo.

Helped a lot