k-means clustering how to indicate what attributes will be used?

caryknoop

How do you indicate what columns to use for k-means clustering. 

 

Assume you have the following attributes:  BadlyDistributedID,  IdentifyingName, Col1, Col2, Col3.

 

Only Col1, Col2 and Col3 should be considered, how to do this?

 

 

MartinLiebig

Hi caryknoop,

 

for all learning task, all regular attributes will be used. Simply set the role of your Name and ID to something using Set Role. You can type anything except regular there. Roles need to be unique.

~Martin

Telcontar120

You can also use a "select attributes" operator if you want to limit clustering to only a subset of the regular attributes in the file, just place it immediately before your clustering operator.

 

andreas_camenzi

Hi Martin

 

Thank you for your help.

 

Do you also know a solution if the attributes are more than 7. We want to display a chart with 2 attributes but need the information of around 10 additional attributes.

 

It is only possible to assign 1 additional attribute per role.

 

Thanks & best regards

Andreas

 

If we just want to include 

MartinLiebig

Hi,

 

you can simply type anything you want instead of "label","id" etc. just type att1, att2, att3 for the desired role.

 

 ~Martin