“Generating different train and test splits is called cross-validation. This helps us have a more reliable estimation of our model's accuracy”