Predicting the Behavior of Loyalty Programs Customers Using Interpretable Patterns Based on Logical Analysis of Data

Aeroplan Inc. ("Aeroplan") , aims to redesign and optimize its loyalty program Aeroplan via a collaboration with Polytechnique Montréal. Customers affiliated with Aeroplan’s program earn miles through their purchases and can exchange these miles for various gifts. It is essential for Aeroplan to predict customers behavior, to define the causes of certain behaviors and to predict the consequences of applying different policies, e.g. offers or gifts value. In this project, we propose to exploit the historical customers database of Aeroplan to predict customer behavior using Logical Analysis of Data (LAD) as an interpretable machine learning technique. We intend to use the LAD generated patterns to design customized marketing plans. This approach should allow the marketing department at Aeroplan to identify the future behavior of customers along with its cause and to target customers with suitable personalized marketing policies. This should help Aeroplan to avoid problems such as churn or drop in usage rate. The prediction accuracy of our model will be compared to traditional machine learning techniques that are known to perform well in predicting customers behaviors. The proposed marketing policies will be tested on a sample of Aeroplan’s customers using an A/B testing approach.

Intern: 
Mohamed Ossama Hassan
Faculty Supervisor: 
Antoine Saucier
Province: 
Quebec
Partner: 
Discipline: 
Program: