A Recommender System using Principal Component Analysis

In this paper we examine the use of a mathematical procedure, called Principal Component Analysis, in Recommender Systems. The resulting filtering algorithm applies PCA on user ratings and demographic data, aiming to improve various aspects of the recommendation process. After a brief introduction to PCA, we provide a discussion of the proposed PCADemog algorithm, along with possible ways of combining it with different sources of filtering data. The experimental part of this work tests distinct parameterizations for PCA-Demog, identifying those with the best performance. Finally, the paper compares their results with those achieved by other filtering approaches, and draws interesting conclusions.