Supervised Machine Learning with Python

Limitations of ALS

We've been using explicit ratings. For example, on Amazon, ratings are between one and five stars. The problem here is that explicit rating systems typically have trouble getting users to rate the items, because it's easier to consume that content than it is to evaluate it from the user side. So, implicit ratings are the inverse of explicit ratings and they can be collected by a system, usually, without the user's awareness. A lot of times that's more favorable, because it doesn't require the user to interact with the system in a secondary sense to explicitly rate items, and we can get more data, which means less sparse data. So, implicit ratings might include the number of listens to a song. There's really well-known ratings ...