It looks interesting to me to know about the variables related to the students performance, so I started to look into the following dataset: https://archive.ics.uci.edu/ml/datasets/Higher+Education+Students+Performance+Evaluation+Dataset

One of the first things I noticed was the nature of the dependent variable.

OUTPUT Grade (0: Fail, 1: DD, 2: DC, 3: CC, 4: CB, 5: BB, 6: BA, 7: AA)

This is obviously an ordinal variable, as the different levels can be easily ranked.

According with the description in the link before, this is a classficiation problem, but the first approach that came to my mind to deal with the ordinallity of the data was to convert it into a regression problem.

Is that approach correct?

In case it is not: How to make my model learn about that sorting/ranking then? How to deal with it in the evaluation phase? For example I see obvious to penalize more an incorrect precition of DD instead of AA, than BA instead of AA.

Any other ideas to deal with ordinallity in a classification problem?

  • 1
    $\begingroup$ I’m used to these things mapping to a continuous variable. F is below 60%. With the rubric you could map it to a continuous, and that’s a different sort of information than an ordinal. $\endgroup$ May 12 at 1:17
  • 2
    $\begingroup$ There are ordinal regression models, such as proportional odds ordinal logistic regression. $\endgroup$
    – Dave
    May 12 at 2:04


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.