Greetings.
The use of holdout data is to see how well a model will generalize and perform on data that was not trained upon. This aids in the prevention of overfitting a model and being subjected to a false sense of accuracy in how well a model is performing.
Hope this helps.
... View more
