Author(s)

P. S. Coelho, G. Lachtermacher & N. Ebecken

Abstract

A comparison of some tree based prediction tools P. S. ~oelho'.~, G. ~achtermacher"~ & N. beck en^ ' ~aculdades Ibmec, Brazil 2 COPPE, Universidade Federal do Rio de Janeiro, Brazil 3 FACC, Universidade Federal do Rio de Janeiro, Brazil Abstract The prediction task (also called regression) uses the attribute values of a given object to infer a response value to a variable. In general, examples of input and output variables are used to construct a model, i.e., the model described the response variable using attribute values. This paper makes a comparative study among several business and academic prediction tools based on tree technology. These techniques (Regression Trees or Models Trees) were developed over the last 20 years and were selected because they have good results according to most common comparison criteria (accuracy, velocity, scalability, robustness and interpretability). Brazilian financial data, obtained from a model that decomposes the stock returns into a specific set of facto

Keywords