Feature Selection: How to select categorical features in a regression problem

Question

I am reviewing information for feature selection based in filter methods. I got info (link1, link2, link3, link4, link5) for:

• Numerical input, numerical output
• Categorical input, categorical output
• Numerical input, categorical output

However, I'm having a hard time finding information on:

• Categorical input, numerical output (categorical features in a regression problem.)

I would be grateful if you could pass me information about it, please, or the name of the function that could carry out this task.

Answer 1

If the categorical input (random) variables are ordinal, then you can use an ordinal encoding to numerical (random) variables and apply the "numerical input, numerical output" schema. Otherwise, if the categorical input variables are nominal, then there is no natural ordering between the values of such variables and better encodings must be exploited to use the same "numerical input, numerical output" approach (such as one-hot encoding).