I've been reading up on feature extraction methods - but the ones I have come across all seem to be numerical. To do feature extraction on a mixed numerical/ categorical dataset are there techniques designed for that? or do you work on categorical/ numerical variables separately?
-
1$\begingroup$ What feature extraction methods do you mean? For instance, I would consider it feature extraction to take a continuous measurement, a categorical measurement, and their interaction (i.e. ANCOVA with an interaction). $\endgroup$– DaveAug 18, 2021 at 17:02
1 Answer
Categorical data is usually converted to numerical by generating dummy variables out of it (OneHotEncoder in Python). These variables are then considered in general as numerical variables and can be used in feature extraction measures.