Currently Foreshadow could not handle classification problems whose target variable is categorical, for example, categories like ">=50k" and "<50k". This change enables the labelencoding on y variables if we are dealing with classification but do nothing if it is a regression problem.
The interface has changed since we have to specify the problem type when initializing the foreshadow object.
The unit test on running classification on the adult_small dataset with tpot showcases such capability.
Description
Currently Foreshadow could not handle classification problems whose target variable is categorical, for example, categories like ">=50k" and "<50k". This change enables the labelencoding on y variables if we are dealing with classification but do nothing if it is a regression problem.
The interface has changed since we have to specify the problem type when initializing the foreshadow object.
The unit test on running classification on the adult_small dataset with tpot showcases such capability.