Suppose that I 've built a ML model predicting when a system (i.e. generator) will break and I have trained it on a specific dataset (train, test). When the model gets deployed, it will get data (input) from an API.
Each time new data come in (input), will it have to get re-trained on the training dataset and then run on the new data to make the predictions (output) or directly run only on the new data? Which means, is there any way to keep in trained in order to avoid re-running it on the training dataset?
Perhaps, my question is kind of ingenious, but I cannot find a clear answer on the web!