I have an alpha that emits insights.
I want to train my XGBOOST/LSTM model using the factors of the alpha (gap, correlation, volume, etc) and the target should be the result of the trade: profit or loss.
(so I can feed the factors to the model later and get a prediction on the likelihood of a profit )

what is the best practice to do this?
I thought about logging the data and train the model with this data - but it's complicated and the logging is limited.

any advice?