CS 69191: Masters Seminar
CS 89191: Doctoral Seminar

Spring 2009


Masters Student Presentation:
Recommender Systems

James Van Heyde


Recommender systems provide ecommerce websites that have enormous product catalogs to provide a better user experience for their customers while at the same time bringing attention to less popular catalog items that would otherwise remain in inventory. With these systems customers are shown more relevant real time recommendations for products that likely appeal to their individual interests. We will examine two methods for recommending items to users, one from Amazon.com and the other from the BellKor submission to the Netflix Prize. The Amazon model uses an item-item recommendation engine while the Netflix Prize model uses a hybrid approach that implements pieces of the item-item along with other more traditional user-user techniques. Both techniques are extremely effective under real world situations in which the user input only represents a very small sample size of the overall data set.