Which domain the problem belongs to? Given a set of products some are classified as cheap and some not. The task is to determine the price range (probablistic) for cheap products ? Supervised classifiers arrive at a function and can classify a new instance but does not provide the range of feature values. This seems like a pattern recognition problem but I am not able to get which class of problems it belongs to.

reading more about this, could this be a rule mining problem?

  • $\begingroup$ like how to find a pattern saying, all products with price <2000 are generally classified as cheap for 90% of the cases. This would be the output after learning over 100s of product features (say for example). $\endgroup$ – user439521 Nov 22 '14 at 18:58
  • $\begingroup$ Doesn't the decision depend on the type of product? A \$1000 car would be cheap but a \$1000 toy car would not. $\endgroup$ – Emre Nov 22 '14 at 21:09
  • $\begingroup$ @Emre yes it does, but for simplicity lets assume all products are from the same category. $\endgroup$ – user439521 Nov 23 '14 at 5:33
  • $\begingroup$ Why don't you just estimate the conditional (on the class) density of the price then? $\endgroup$ – Emre Nov 23 '14 at 6:28
  • $\begingroup$ Do you have more than just the price for each item and would an item labeled “cheap” receive that label again if the labeling process were repeated? (I.e. is there a eater agreement problem lumped in with your task?) $\endgroup$ – Wayne Jul 9 '19 at 14:16

Based on your comment, it seems you are searching for a decision tree algorithm here with 1 split per node. See the C4.5 algorithm

The idea you describe can be done with other classification algorithms, but you would have to fix all other features per instance, and then manually vary one feature at a time until you cross the decision boundary.

| cite | improve this answer | |
  • 1
    $\begingroup$ reading more about this, could this be a rule mining problem? $\endgroup$ – user439521 Nov 23 '14 at 8:00

Your Answer

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

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