When you start learning and creating machine learning models, you might run into a few different a few different ways used to describe how your model is running or being used.

Training is when you take your data and try to make your model learn how to solve your problem. Evaluation is when you take your data and you try to see how well your model is performing currently. During evaluation you still run all of your metrics, you just aren’t updating the model.

Prediction is when you take the data and you try to make a guess as what the solution is. You don’t run any metrics and you aren’t updating the model, but instead you are just taking in the input and trying to predict the answer. There will often be some kind of mapping that occurs here, especially in classification problems where you are trying to say that greater than .5 is a cat or that anything in this cluster is French.