multiple labels not sees as labels

wasperen

Hi All,

I understand that a learner (for instance Neural Net) tries to learn the function to get from "regular" attributes to "label" attributes. When I change the role of an attribute from "regular" to "label" and then the next one, the first "label" just disappears. That seems to be normal behavior.

So I created a loop that assigns roles "label1", "label2", etc to my target attributes. That works well and I can see the attributes that are my functions input as "regular" and my target outputs as "labelN" (where N is the number of the label).

So far so good.

But now the Neural Network learner complains that it does not see any "label" attributes in the training set...

I have tried "label_Attribute Name" and "label Attribute Name" and "label_N" but to no avail.

Should the learner not recognize any role that starts with "label" as the target outputs?

Thanks,
Willem

land

Hi Willem,
most of the learner are not capable of predicting more than one label at a time. So what you can do is to generate one model per label. You can iterate over your lables using a loop attributes operator and setting the current labelN to the role label. Then build a model on this label, and for example save it as ModelN.

Greetings,
  Sebastian

wasperen

Sounds reasonable! Thanks. I'll try it out this weekend.

Regards,
Willem