Decision trees (C&RT)

Decision trees are tree-like graphs or models that help to identify strategies to reach decisions about specific goals. This is a useful method to display an algorithm and choose between several paths of subsection (Wikipedia, 2011). A decision tree consists of nodes and splits. The tree starts with a single node including all training data. The initial node split into two new nodes by using a predictor variable. And then the two nodes split into two new splits by using the second and third predictor variables, and this continues until a terminal node. A terminal node is one where no more splits are made (StatSoft, Inc., 2011).

A decision tree may include a large number of nodes but it is not hard to evaluate and reach a decision because it is very simple and straightforward for interpretation. Generally the decision tree models give good accuracy rate, but sometimes can be minor challenges. At these times, a researcher need to find the right sized tree, and this requires time and experience (StatSoft, Inc., 2011).

1-      StatSoft, Inc. (2011). Electronic Statistics Textbook. Tulsa, OK: StatSoft. http://www.statsoft.com/textbook/