Problem

Streaming platforms want their users to consume content in order to gain fidelity. For that, it's important that they enjoy the new content they watch. That's why these kind of platforms want to understand what kind of content each users like the most. How can we deliver that content when tastes are so different?

‍

Solution

Well, at first it may seem that your taste cannot be linked with anyone elses'. Minds are so different form each other that trying to generalize could lead to mistake. Undeniable indeed.

However, there is a solution to this problem.

It has been proved that there's people that end watching the same movies, or that they both end up watching a fair amount of them. That being the case, it's an interesting strategy to find people with a high coincidence in movie/series taste, and recommend movies that one of them hasn't watched, to the other person and viceversa.

That's how the algorithm works. The only problem is how to manage big amounts of data, that is, large numbers of users.

That's where SVD enters the game. Signular Value Decomposition reduces a large matrix into a smaller one determined by the number of eigenvalues chosen. Basically an ideal application for a case such as this one.