A hybrid sequential system for inflation forecasting

Abstract

Forecasting price indexes of the economy has received an attention from scholars and policy makers due to its significant effect on various sectors and markets. The stability of economy is at risk if inflation is not properly checked through models and macroeconomic studies, therefore, forecasting inflation is an important task for the formulation of policies in governments and companies. In this sense, this research work attempts to model inflation rate to Brazil, United State of America and Japan. The literature, in the field of time series, indicates the combination of linear and non-linear models to model inflation. In this paper, a hybrid ARIMA-MLP system has been proposed to map linear and nonlinear patterns. This is explored using a hybrid evolutionary system consisting of a simple exponential filter, linear ARIMA and autoregressive (AR) models and a Multilayer Perceptron model. In addition, it was also implemented exponential smoothing models (ETS), Qunatile Regression (QR) and Support Vector Machines. The experimental results show that the hybrid evolutionary system presented promising results in the prediction domain.