A Novel Recommender System for Healthy Grocery Shopping

Abstract

Given here is an anonymized dataset of online grocery purchases from users; we present a recommender system framework to predict future purchases. We describe the method of constructing a utility matrix to run a collaborative filtering algorithm to pair similar and dissimilar users and ultimately provide recommendations. Given those recommendations, we further our analysis by proposing a method using natural language processing to determine the nutritional value of a food product to further improve recommendations. The results provide recommendations for the healthiest options based on historical purchase data.