Predicting Purchasing Behavior at PriceMart (B)

Supplements the (A) case. In this case, Wehunt and Morse are concerned about the logistic regression model overfitting to the training data, so they explore two methods for reducing the sensitivity of the model to the data by regularizing the coefficients of the logistic regression.
Wehunt and Morse then compare the models and select the model most effective at correctly classifying households as expecting. Students explore the relationship between the model’s confusion matrix, which organizes the model’s correct and incorrect classifications, the cutoff point on the curve that matches true positives and true negatives, and the payoff matrix Wehunt and Morse construct. Students can then follow the link directly from their model to their marketing strategy.

Technical topics covered:
o Ridge logistic regression (or L2 regularization) as a modelling technique
o Lasso logistic regression (or L1 regularization) as a modelling technique
o Comparing models, thinking about coefficients, and selecting model for deployment
o Evaluating model output; ROC curve, cutoff point, confusion matrix; payoff matrix as a framework for utilizing the model to carry out marketing strategy