Machine learning has methods to prevent overfitting in both training and testing data. It is not true that machine learning always overfits; it depends on how it is used. When dealing with unstructured data, machine learning is the best option. Classical statistics should be learned before machine learning to focus on preventing overfitting.