I am looking for a multiclass classifier that can handle repeated measures. Specifically, each of my subjects appears multiple times with the same number of n classes. Now I would like to fit a classifier to this dataset to get an intuition for the feature-label relationship (that is, which features are the most associated with which classes?). Is there a multiclass classifier that can take this into account?


My problem is a multi-class problem, not a multi-label problem. I did not know that 'multilabel' and 'multiclass' describe different problems, so I changed the wording.

enter image description here

  • $\begingroup$ All your data points have the same labels? I can see both subj_id 1 and 2 having the labels 1, 2 and 3. Or I’m missing something? $\endgroup$
    – rivu
    Commented Oct 21, 2021 at 14:23
  • $\begingroup$ Yes, that's correct. I am interested in the feature label relationship while accounting for possible within-subject correlations. $\endgroup$ Commented Oct 21, 2021 at 14:32
  • $\begingroup$ If every point has the same class labels [1, 2, 3], this is not a multi label classification problem. For a feature f, every value it can take produces a class label l. In other words, the class label is not an RV. So I don’t know if a correlation can even be computed. $\endgroup$
    – rivu
    Commented Oct 21, 2021 at 14:44
  • $\begingroup$ Imagine training the classifier only using the features and the y label and further imagine you wouldn't know the subject id. Then someone comes and gives you this additional information. Then I guess one could argue that feature sets within each subjects are more similar to each other than they are between subjects? In other words, that there is a nested structure in the feature set? $\endgroup$ Commented Oct 21, 2021 at 15:06
  • $\begingroup$ Do all your subjects have the same labels in the same order? e.g. subject_1_labels = [1, 2, 3], then subject_2_labels = [1, 2, 3], subject_n_labels = [1, 2, 3]. Then I don't think there will be any predictive method informative here since the labels are all the same across subject and also have the same ordering as well. $\endgroup$
    – Kirk Walla
    Commented Oct 22, 2021 at 15:25

1 Answer 1


This might not be a complete answer but as a first step I would just fit a decision tree and then plot the feature importance to get a better understanding regarding the relationship between $X$ and $Y$. For more info regarding weather your case is a multi-label or a multi-task setting you can look here and decide accordingly as to how to proceed.

  • $\begingroup$ Sorry, I just realized that I mistakenly used the term 'multilabel' instead of 'multiclass'. Made an edit to my question. $\endgroup$ Commented Nov 9, 2021 at 18:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.