Decision trees are typically schematic, tree-shaped diagrams used to show a statistical probability. Classification and regression trees (CART) are one of the most well-established supervised learning techniques. CART works by repeatedly finding the best feature to split the data into subsets.
The partition improves the isolation of the label with each split. Decision trees can be used for either classification, for example, to determine the category of an observation (that is, default or nondefault), or for prediction, for example, to estimate a numeric value (that is, the loss given default).