Evaluating the combined forecasts of the dynamic factor model and the artificial neural network model using linear and nonlinear combining methods

Abstract

The paper evaluates the advantages of combined forecasts from the dynamic factor model (DFM) and the artificial neural networks (ANN). The analysis was based on three financial variables namely the Johannesburg Stock Exchange Return Index, Government Bond Return Index and the Rand/Dollar Exchange Rate in South Africa. The forecasts were based on the out-of-sample period from January 2006 to December 2011. Compared to benchmark autoregressive (AR) models, both the DFM and ANN offer more accurate forecasts with reduced root-mean-square error (RMSE) of around 2–12 % for all variables and over all forecasting horizons. The ANN as a nonlinear combining method outperforms all linear combining methods for all variables and over all forecasting horizons. The results suggest that the ANN combining method can be used as an alternative to linear combining methods to achieve greater forecasting accuracy. The ANN combining method produces out-of-sample forecasts that are substantially more accurate with a sizeable reduction in RMSE of both the AR benchmark model and the best individual forecasting model. We attribute the superiority of the ANN combining method to its ability to capture any existing nonlinear relationship between the individual forecasts and the actual forecasting values.