Best practices for one-vs-rest predictive modeling

[Max]

There is cost-sensitive learning that could help. It's pretty straightforward. brulee or keras can do that.