What to Watch: An Examination of Matrix Completion Techniques Used in the Netflix Prize

Abstract

With the rise of big data comes the problem of how to properly leverage it into business insights. One area of concern is how to effectively predict customer sentiment towards products. Using matrix completion it is possible to take an incomplete matrix of users and their ratings of products and extrapolate the data to suggest new products. This problem gained considerable notoriety in the past decade with the Netflix Prize competition. However, many current methods are either over specialized by dataset, produce only theoretical results, or are overly simple. The purpose of this paper is to look at current techniques and identify an optimized method that can work on a variety of data sources.