Predict which free-trial users of a course app will buy, and tune the threshold for the sales team.
New to this? See the worked example firstLogistic regression predicts a yes or no answer, such as whether a free-trial user will buy, and gives the probability behind it. In these questions you build the answer column, train a scikit-learn model, and judge it with a confusion matrix, precision, recall and ROC-AUC.
Interviews often ask why accuracy is not enough for rare outcomes, how to move the decision threshold, and what a coefficient means for the odds.
If you are new to it, read the worked example first, then solve the questions in order: the early ones explore the data and train a first model, the middle ones test it on rows it has not seen, and the last ones tune it and turn the result into a business decision.