We have deployed a machine learning model to production around 1 year ago. I would like to somehow estimate performance of my model (binary classifier), but we only get ground truth about 9 months after the prediction is made. I've noticed that the business impact after the model was deployed has degraded significantly, so I need to be able to estimate current performance without having to wait for ground truth.
Are the any methods / products / libraries for that, or am I out of luck?
