Previsão da inflação brasileira através de modelos de aprendizado de máquina – LASSO

Abstract

The significant impact of artificial intelligence on various sectors of a country is well known. In turn, this phenomenon has also extended to the economy and politics. Therefore, this study aims to shape a scenario in which artificial intelligence, through machine learning, is used as a complementary tool in the process of forecasting macroeconomic variables, such as inflation measured by the IPCA, and consequently assist the agents involved in decision-making. Under this perspective, the LASSO method will be employed as a variable penalization model. Thus, the selected algorithm eliminates variables considered less impactful for the variable of interest. To highlight the efficiency of this technique, performance measures such as MAE and RMSE will be applied, revealing the model’s relevant accuracy. Finally, in order to verify the effectiveness of LASSO, this algorithm must be compared with benchmark models.