I need some machine learning vocabulary advice.
What do you call input data that a trained model gets in production?
I know that labelled or unlabelled data it gets during training is called training set or training data.
When it's ready, after initial training, you deploy it in production to solve real problems. How do you call the input data that the model receives at that point?
I think you're talking about production data. This is typically real data that you generate or acquire somehow, and it's the data you expect your trained model to classify or use to make predictions.
You mention that this is after initial training. I'm assuming you've already used test data to measure your model's accuracy, and this is in fact now in production.
It is subjective what you are going to call your "Unseen Data", and it would not matter IMHO. You may call it production data as it is suggested above, or even unseen data or test data. In the Machine Learning Glossary by Google Developers, you find all standard definitions. What I would personally prefer to call is either test data or hold-out data:
Examples intentionally not used ("held out") during training. The validation dataset and test dataset are examples of holdout data. Holdout data helps evaluate your model's ability to generalize to data other than the data it was trained on. The loss on the holdout set provides a better estimate of the loss on an unseen dataset than does the loss on the training set.
After all it is the data that is held out and model has not seen, whether intentionally or naturally, that model need to do the predictions on. I have to admit that both these names imply that at present you may own such data, and it is kind of contradictory to the fact that such data is not avaialble yet; but I would still use one of these terms as long as myself and my audience is fully aware of it.
