0
$\begingroup$

Does anyone know of any packages that implement multilabel feature selection algorithms? There are many papers introducing algorithms, but few if any seem to be implemented publicly. I have been able to adapt some standard feature selection algorithms for multilabel, but I'd really like to employ some specifically designed for the task

$\endgroup$
3
  • $\begingroup$ Welcome to DataScienceSE. Multilabel classification is essentially the same as training N independent binary classifiers, with N the number of possible labels. So in this case I think that I would do feature selection for each independent binary classifier, in order to find the optimal set of features for every label. $\endgroup$
    – Erwan
    Aug 26 at 21:32
  • $\begingroup$ Thanks @Erwan! Thats what I'm currently doing. I was hoping to use some of the algorithms specifically designed for multilabel classification, because the main difference between multilabel and binary classification is the potential for dependence between labels. Unfortunately, feature selection on N binary classifiers would miss that. It seems that most of the multilabel specific algorithms haven't been implemented publicly though $\endgroup$
    – tensormoby
    Aug 27 at 13:47
  • $\begingroup$ Ok I see. Genetic feature selection could be an option, but it requires training/testing many times so it depends how heavy the training process is. $\endgroup$
    – Erwan
    Aug 27 at 20:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.