TBF, the major alternative to scikit-learn, doing it in R, would have you doing model <- glm(formula, family=binomial(link=logit)); predict(model, type="respons

TBF, the major alternative to scikit-learn, doing it in R, would have you doing model <- glm(formula, family=binomial(link=logit)); predict(model, type="response"). R also doesn't make bootstrapping intuitive (there are a number of fairly easy to use packages, but nothing that's as canned as you'd hope it would be). Granted I still use R for ML stuff, and I concede that silently regularizing regressions, and only being able to avoid doing so by hacking the penalty parameter, is terrifying.