Applying Multi-label Transformation in Rapidminer?

Question

I am working on text categorization in rapid miner and require to implement a problem transformation method to convert multi-label data set into single label i.e. Label Power set etc but couldn't find one in Rapid miner, i am sure i am missing something or may be Rapid miner has provided them with another name or something ?

1) I searched and found "Polynomial By Binomial" operator for Rapidminer which i think is using Binary Relevance internally for problem transformation but how can i apply others i.e. Label Power set or Classifier Chains ?

2) Secondly SVM (Learner) inside "Polynomial By Binomial" operator is applied K(Number of classes)times and combines 'K' Models into a single model but it would still classify a multi-label (multiple labels) example as a single label (one label) example, How can i get the multiple labels associate with an example ?

3) Do i have to store each model generated inside "Polynomial By Binomial" and then apply each on testing data to find out the multiple labels associate with an example ?

I am new to rapid miner so ignore my mistake

Thanks in Advance ...

Answer 1

Polynomial to Bionomial is not the way you want to go.

This operator performs something like XvsAll. This enables you to solve multiclass problems with a learner only capable doing binomial classification.

For your problem: Would it to transform your table like this:

before: ID Label 1 A|B|C 2 B|C to ID Label 1 A 2 B 3 C 4 B 5 C

The tricky thing for this is how to calculate the performance. But i think once this is clear a combination of recall/remember/remove duplicates and join will do it.