I trained a StateSpaceForecast algorithm and saved it with the fit command.
My goal is now to make predictions on (near) real time data. My idea was to use the holdback variable which I would set to a positive value, and compare the prediction for (let's say) 10 minutes ago, with the actual value 10 minutes ago and if there is a difference, I sent an alert.
I want to do this with the apply command. I am unsure however, whether or not the apply command retrains my model (including new datapoints in the training) or whether it just applies the StateSpaceForecast with exactly the parameters specified in the previously done fitting?