"Data Importation: Set Role Issue"

alourenco

Hi all!

I am trying to use the dataset files from the UCI archive in ARFF format in RapidMiner, but I am facing a huge problem:

most of the times, the set role operator does not pick the dataset from the ReadARFF operator and so I am not able to define the label attribute.

This is weird because now and then, all of sudden, it does pick it and provides me the list of attributes from which I can choose from.
I have tried to insert the operator right after the readARFF, then after having a classifier associated (through a quick fix error), ... and nothing.

Could someone help me with this please?

Best,
Ana

Marco_Boeck

Hi,

the "Set Role" operator tries to use the meta data of the input to determine which attributes to present to the user. However, sometimes the input is not available without actually executing the process (which is not done automatically as it might take a long time) as in your case. There are two ways around the problem:
1) Enter the attribute name anyway. You are not restricted to the choices offered, and if the attribute name is correct at runtime, you can ignore the potential problem RapidMiner points out.
2) Import your data and save it to the repository, then load it from there. Data stored in the repository and retrieved from it has meta data attached right away.
3) Active the meta data propagation (the button with the paper clip icon in the upper right corner), add a breakpoint after the "Read ARFF" operator (rightclick it, select "insert breakpoint after") and then execute the process until the breakpoint. Then switch back to the design perspective, and now you should have the attribute choices in the "Set Role" operator.

Regards,
Marco