I have a machine learning Random Forest model that predicts a certain variable. It's implemented with scikit learn and it works fine.
Now, assuming that the prediction relates to month 1, I need a new model to predict the variable of month 2, 3, and so on until 12 months. So, what I need is to predict the variable for month N using the prediction of month N - 1.
To do this manually, I would have to process each month in a loop, adding the previous prediction to the historic data, and retrain the model.
The problem is that retraining takes a long time, and I would have to do it 12 times.
Is there a way to "add" to the trained data the new values, without retraining? Or maybe my approach should be different?