The train, test splits, and cross-validation sets are a fundamental concept in machine learning. This is one of the areas where a pure statistical approach differs materially from the machine learning approach. Whereas in a statistical modeling task, one may perform regressions, parametric/non-parametric tests, and apply other methods, in machine learning, the algorithmic approach is supplemented with an element of iterative assessment of the results being produced and subsequent improvisation of the model with each iteration.