# Implicit Matrix Factorization for Recommendations: Are ratings=0 required?

I am working on building an Implicit Matrix Factorization Recommendation System. I have data that contains ratings between 0 and 28, based on the frequency of views (User A viewed Product XYZ a total of 8 times, thus has a score of 8). I was reading through the Hu et al. paper and from my reading, it appears that I also need to include "zero ratings", meaning an item that a user has not viewed is logged as a score of 0. Is this understanding, that I must include "zero ratings" correct?

That is correct. The response $p$ is 0 when there are no views and 1 when there is at least one view. A weighting variable is added called the confidence $c$. This confidence equals 1 when there are no views and increases monotonically with the number of views. The model includes all of the "ratings", but the confidence gives less weight when there are no views.