Generating realistic customer purchase baskets using Generative Adversarial Networks (GAN)

This project uses the purchase history data from loyalty member cards at the basket level (i.e., all items bought by a customer during a particular trip) for customers under the loyalty program of a chain drug store to develop a model that can generate a realistic future “customer shopping list” (i.e., customer baskets) using the novel machine learning technique of Generative Adversarial Networks (GAN). Specifically, the goal of this project is to use the above data to build a simulator that would predict for a given customer, multiple possible future baskets of items during his/her next four trips (one trip per week) conditioned on the previous weeks’ baskets. This project has significant potential applications to allow retailers to better predict demand and revenue potential at individual customer level including better supply chain and inventory management as well as more personalized promotion planning.

Faculty Supervisor:

Saibal Ray

Student:

Thang DOAN

Partner:

Rubikloud Technologies Inc.

Discipline:

Business

Sector:

Finance, insurance and business

University:

McGill University