Kirk Boyer, University of North Carolina at Chapel Hill


Content-driven Association Rules for Recommender Systems

Abstract: Businesses like Amazon.com and Netflix are increasingly interested in algorithmically providing personalized recommendations of items (movies, books, etc.) to their customers based on feedback given about previously purchased or used items. This project applies the data-mining concept of association rules to a combination of data involving users' ratings of items and item-characteristic (content-based) data to provide a dynamic strategy for generating recommendations. Part of the goal of this project is to generate reliable and accurate recommendations while avoiding complete reliance on arbitrary rating scheme such as a scale of 1 to 5 stars, an almost ubiquitous encumbrance on current algorithms.

Mentor: Amy Langville (College of Charleston)