Using Decision Trees for Prediction of US Economic Recessions

Abstract

This paper examines the efficiency of decision trees on US economic crisis periods. Many other studies examined various approaches, like noise-to-ratio models, discrete choice models, neural networks, fuzzy logic and neuro-fuzzy systems among others. Two approaches are applied. The first is a discrete choice using both binary Logistic and Probit regressions, while the second approach is a decision tree. We propose the use of three models for two reasons. Firstly, with discrete choice models we can examine and investigate the effects and the magnitude of the inputs or the independent variables. Secondly, decision trees are able to account for much more inputs in the model, outreaching the problems of variables significance.