ENSEMBLE OF MACHINE LEARNING APPLIED TO ECONOMIC CYCLES ANALYSIS: A COMPARATIVE STUDY USING ANTECEDENT MACROECONOMIC INDICATORS FOR BRAZILIAN GDP PREDICTION CLASSIFICATION

Resumo

This work proposes a comparative study between several machine learning techniques, applied in the analysis of the phases of the Brazilian economic cycle. To this end, several macroeconomic indicators were used to build a model that was able to identify and predict the turning points of the economic cycle, such as the beginning of a recession or a recovery. The discretization of the variables proved to be decisive in the quality of the classification process, due to the diversity of the data and the non-linear nature of the analyzed phenomenon. The different techniques used reinforce a dilemma, because usually the best results come from very abstract methods, making it difficult to interpret the steps and their causes.