HeatWave Release Notes
MySQL 8.0.33 introduces recommendation models which can recommend the following:
The rating that a user will give to an item.
The top K users that will like an item.
The top K items that a user will like.
Recommendation models include matrix factorization models, and use Surprise algorithms, see: Surprise.
MySQL 8.1.0 adds further options for recommendation models which can now also recommend the following:
The top K items similar to another item.
The top K users similar to another user.
MySQL 8.2.0 introduces recommendation models for implicit feedback. When a user interacts with an item, the implication is that they prefer it to an item that they do not interact with. Implicit feedback uses BPR: Bayesian Personalized Ranking from Implicit Feedback which is a matrix factorization model that ranks user-item pairs.
Recommendation models that use explicit feedback learn and recommend ratings for users and items. Recommendation models that use implicit feedback learn and recommend rankings for users and items.
Ratings from explicit feedback are specific values, and the higher the value, the better the rating.
Rankings are a comparative measure, and the lower the value, the better the ranking. Because A is better than B, the ranking for A has a lower value than the ranking for B. HeatWave AutoML derives rankings based on ratings from implicit feedback, for all ratings that are at or above the feedback threshold.
Recommendation models can now repeat existing interactions from the training table.