Collaborative filtering has been widely and successfully applied to many recommendation problems in which items need to be recommended to users based only on the result of past recommendations. Such models attempt to represent both the uniqueness of each user and item, as well as the underlying correlation between groups of users and items. An underlying correlation among users is also present in many situations of strategy recommendation. However, strategies can be better viewed as a set of choices which can be continuous (e.g., how much to invest in each sector of a company). We intend to modify standard collaborative filtering approaches to recommendation systems to extend them to strategy recommendation, proposing a theoretical model and recommendation algorithm that minimizes regret.

“After having a great experience as a UROP student for one semester, I decided to have even more commitment and exposure to research through SuperUROP. My previous UROP and my coursework were focused on data science, and this will be a great opportunity to apply what I’ve learned. I expect to learn more about the creative process of designing new mathematical models for real world problems, and I’m thrilled to contribute to the field I love.”