Date of this Version
Arens, Jack; Carrigan, Joseph; Huck, Ray; Neupane, Nischal; Sullivan, Jacob. The Buckle: Product Recommendation Engine. Undergraduate Honors Thesis. The University of Nebraska Lincoln. 2021.
This project is the culmination of a few different machine learning model creation techniques and discussions with The Buckle’s internal stakeholders. Our final product is a neural-network based machine learning model that recommends products to The Buckle’s customers through their various marketing channels such as email marketing or directly on their website. To create our model, we used a dataset that included products in hundreds of thousands of purchases by The Buckle’s customers. Using this data, our model creates a large multidimensional space with each of The Buckle’s products. This multidimensional space is then reduced down to 50 dimensions where related products are located closer together. The model recommends products by taking in a customer, creating a point in this 50 dimensional space based on the customer’s previous purchases, and recommends items that are the closest to this point. The benefit to this model over any of our previous models is that The Buckle can filter what type of item is recommended by availability, demographic, item type, etc. As a result of this project, we have given The Buckle an internal tool that can replace their third-party product recommending software. This tool can also be expanded upon internally to fulfill more use cases than the third-party software can.