Using Regression Trees in Predictive Modelling

Abstract

Tasks where the value of an unknown parameter has to be estimated are typical in data mining projects. The solution of this kind of problems requires the creating of a predictive model. There are several methods for the solution of this type of tasks, e. g. the decision trees and the regression. In the paper an algorithm is demonstrated which combines the properties of the two techniques. Tests on different datasets show the efficiency of the method compared to the results which were provided by the traditional regression and decision tree algorithms.