Stock returns forecasting for financial, food, industrial and services companies using neural networks and arima-garch models

Abstract

The main purpose of this work is realize stock returns forecasting for financial, food, industrial and services companies using feedforward neural networks trained with Levenberg-Marquardt algorithm and Arima-Garch models. In each area two time series was selected from Economatica. To the financial area, Bradesco and Itaú was analyzed, Perdigão and Sadia in the food sector, Marcopolo and Gerdau in the industrial area, finally Pão de Açúcar and Lojas Americanas in the services. The forecasting generated by the two techniques had similar performance implying no significant differences between them.

Time series; Forecasting; Levenberg-Marquardt algorithm; Neural networks; Arima-Garch