Decision tree for classification and forecasting.

Classification and forecasting are two recurrent words in business world nowadays. Classification aims at predicting the future class while forecasting aims at predicting the future value of a system that is intrinsically uncertain. This chapter briefly presents the main methods, focusing on the decision tree (DT) methodology. A DT is a hierarchical collection of rules that describe how to divide a large collection of units into successively smaller groups of units. In the Recursive Partitioning Steps a choice of the criteria to split the nodes is given. The prediction value of the leaf depends on the target variable Y. In fact, if Y is a numeric variable, then the prediction value of the leaf is the average mean of the variable Y. Regression trees are DTs where the target variable Y is a numerical variable. The chapter presents a binary recursive partitioning process.