I haven't seen that adding macro features into an AI model makes any difference. I might be using them wrong, or the machine learning algorithms can't handle them well.
So, I wonder if there is any technique to train a model in different market regimes manually. For example, let's say I have a macro indicator for bear/bull markets, and I want to train two models for the two different regimes. Then, I would use a conditional node in my ranking system to switch between the two trained models.
You can do this by simply using in-universe screening.
Also, have you tried macro-related betas?
No you can't train a model with periods of an empty dataset.
IsTicker
We don't have direct support for macro features yet. Macro features are treated like any other feature which means they are normalized using the same preprocessor as stock features. This is definitely wrong with the Rank preprocessor. Maybe a bit better with z-score, but not ideal. It's one of the things that got the most votes here AI PRE LAUNCH: Feedback Roundup
Another one that got votes is support for Categorical features. Once we have categorical features support you could upload your own bull/bear indicator. The idea being that the model will adjust automatically.
Until we have direct support for macro and categorical it's probably a waste of time to try.