Imagine that you must analyse a medical dataset in which a number of

patients are classified as ill or healthy. Imagine that the doctors who classified

the data examples in these data were not very knowledgeable, and they made

many mistakes. This means that many healthy people were classified as ill

and the other way round. Your goal is to decide whether a small or a large

decision tree would be suitable for this dataset to maximise classification

quality on new patients. Would you tune the parameters of the decision tree

learning algorithm to have large or small decision trees? Justify your answer

using technical terms.