Singular Value Decomposition in Recommender Systems
Many websites nowadays such as Amazon, Ebay have used different kinds of
Recommender Systems to predict ratings of items from their clients, so that they could suggest
which items are more likely to be purchased. Nevertheless, problems arise as a data set can be
both too large and too sparse to predict ratings. Hence, the size of a data set and its information
sufficiency should be taken into consideration in order to make accurate predictions efficiently.
In this project, we examine a method of matrix factorization called Singular Value
Decomposition to approach the aforementioned problems by applying the method to real data
that we retrieve from Grouplens.