I was comparing both R and Python implementations of various models,
such as Generalized Boosted Regression and Generalized Linear Models, and I was wondering why in R it is usual to see the
formula parameter and in Scikit-lean it is not.
The question is why was scikit designed this way. Only a few people can factually answer that question. I have my opinion, but that is all that it is.
However formulas can be used with scikit or statsmodels or other packages. Patsy gives the ability. This can be used with scikit as the output of Patsy functions a lot like numpy arrays. An example is here.
The choice to use formulas or work with matrices would be up to the individual.