I am not quite sure about the point of this? Granted, I do understand that in-game item recommendation has the ability to increase revenue. However, it seems like the API is rather limited and inflexible.
What advantage does this system have over having in-game machine learning library like DataPredict to perform in-game item recommendations? All for I know is that the DataPredict library allows you to choose models and data of your choice like taking advantage of players’ data (like current health, number of items, rank and etc.) and use that to create game-specific recommendations.
It doesn’t seem like the Product Intelligence API is even transparent about on what kind of ML it uses. Is it reinforcement learning? Classification? Regression? Clustering? Classic recommendation system using matrix factorization?
Might as well stick with machine learning libraries and have more personalized recommendation that fits with the games’ gameplay.
For anyone who wants to see how you can implement custom ML recommendation systems and other game-related use cases, you can have a look at DataPredict’s “High Value Project Tutorials” here.