Chapter 10: Tree-based classifiers
Daniela Calvetti, Erkki Somersalo · Society for Industrial and Applied Mathematics eBooks · 2020
It is a common practice in decision theory to use branching algorithms consisting of simple yes/no questions, the answer determining the next question until the algorithm arrives at a decision. Branching algorithms are particularly appealing in fields like medical diagnostics where the idea of narrowing down the decision by exclusion is part of everyday practice. This constitutes a promising starting point for the development of automatic smart systems and artificial intelligence. The branching process provides a basis for a tree structure that turns out to be a useful tool in data science for, e.g., classification. In this chapter we discuss supervised learning using tree structures for classification purposes. These methods are often referred to by the acronym CART, which stands for classification and regression tree. Regression trees are briefly discussed at the end of the chapter.