1
$\begingroup$

I have many categorical features that are non-nominal and also continuous features with continuous output. Some of the categorical features are binary and others have 10+ classes. How would you go about building a model to make continuous predications? I was thinking of using some sort of decision tree to learn on the both types of data or maybe have two separate classifiers for each data type and combine the predictions. Should a neural net be used for this?

$\endgroup$
1
  • $\begingroup$ What do you mean that your categorical variables are non-nominal? Do you mean there is an order to them like good/medium/short or tall/grande/venti? (That second one is a Starbucks joke.) $\endgroup$
    – Dave
    Aug 27, 2023 at 2:48

1 Answer 1

0
$\begingroup$

Consider effects coding your categorical variable. You will estimate how much of an effect each category has on the outcome and use that in the model instead of the category itself. There is an R package called vtreat that will do this for you. For more details look at this blog post.

$\endgroup$

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.