![IconGems-Print](images/IconGems-Print.png)
445 – Contributed Poster Presentations: Section on Statistical Learning and Data Science
Clustering for Personalized Preference Prediction
Fan Yang
University of Minnesota
Xiaotong Shen
University of Minnesota
We build a model to allow personalized prediction for different individuals on a large amount of items based on both user features and item features, as in a recommender system. User and item "preferences" are clustered through supervised learning by modeling the observed response with a gaussian distributed regression model. Besides mean parameters, correlation structure of the response variable is also modeled. Fusion type penalties are applied to identify similar users and items. Simulation results show our model performs better than the popular matrix decomposition methods.