Recommendations on grocery shopping: customer or product similarities?

Online grocery shopping becomes more popular every day and benefits both customer and grocery store. To increase revenue, grocery stores recommend products to customers to add to their online baskets. Such recommendations can be either item- or consumer-based. In this research I investigate recommender systems based on the Collaborative Filtering algorithm. First, I replicate part of the paper of Li, Dias, Jarman, El-Deredy and Lisboa (2009), which investigates different standard item-based Collaborative Filtering models and introduces a new personalised recommender system, which outperforms the standard methods. Second, I investigate a user-based Collaborative Filtering model and compare these results to results of the item-based models. The results of the replication do not entirely agree with the results found by Li et al. (2009), which is mainly due to the higher number of items used in the replication. Despite the challenges of the user-based Collaborative Filtering model, the performance of the user-based model is similar to the performance of some of the item-based models, but does not outperform the item-based models. To generate all the results I have used a real-world dataset from the grocery store Ta-Feng.