take a training dataset and split it into 2 parts:
- train set
- test set
training the model can be done on the train set evaluating the model can be done using the test set
the purpose of the split is to reserve data points for testing that was not learned in training