A Market Basket Analysis Product Recommendation

A System for Online Retailers

In the fall of 2019, I was a part of a student group that designed a model to predict future customer purchases given what they have bought in the past. We developed a Market Basket Analysis Product Recommendation System that was able to predict items that would be bought together in the future.

Using RStudio, we first cleaned the data we had from the Kaggle source. Our two association models were split into customers who have reordered products before, and customers who have not. The result of our apriori algorithm resulted in bananas being the most frequently purchased item. The rest of the results were also fresh produce items, with bananas and organic strawberries that had the highest association in frequency.

Overall, we realized the importance of a properly set up predictive model given the class foundation we were given by Professor Lanham. The business conclusion was the importance of fresh produce to returning customers, and the store could benefit from this knowledge by offering frequently purchased items together for the customers. It also holds the potential to increase sales and revenue, and offer promotions for both the customer and the retailer’s benefit.

We presented our final results at the Purdue University Undergraduate Research Expo in the winter of 2019!

It was a great experience presenting our work on our poster, and getting to see all of the innovative research ideas being set forth by my peers.