Black-Box models are actually more explainable than a Logistic Regression

In the following, we will show that not only is there no need to choose between power and explainability, but that more powerful models are even more explainable than the shallower ones. By way of illustration, we will be using one of the most well-known datasets: the iconic Titanic dataset. We have a bunch of variables about Titanic passengers, and we want to predict how likely each passenger is to survive. For what concerns classification problems, Logistic Regression is often taken as the baseline. After having one-hot encoded the qualitative features (Ticket class, Passenger sex and Port of embarkation), we fit a plain Logistic Regression on the training data.