Oh yes, we just include all the possible nonlinear effects and pick the ones which look meaningful [0][1] Then we go back and perturb the model to do some hand

Oh yes, we just include all the possible nonlinear effects and pick the ones which look meaningful [0][1] Then we go back and perturb the model to do some hand wavey guessing at real actionable insights but hey clients are impressed by overfitting so who the fuck cares as long as the money is coming in and the press releases are going out!!!! [0] https://www.featuretools.com/ https://www.featuretools.com/ [1] https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.PolynomialFeatures.html https://scikit-learn.org/stable/modules/generated/sklearn.pr...