A factor — Artificial neural network model for time series forecasting: The case of South Africa

In this paper, the factor models (FMs) are integrated with the ANN model to produce a new hybrid method which we refer to as the Factor Artificial Neural Network (FANN) to improve the time series forecasting performance of the artificial neural networks. The empirical results of the in sample and out of sample forecasts indicate that the proposed FANN model is an effective way to improve forecasting accuracy over the dynamic factor Model (DFM), the ANN and the AR benchmark model. When we compare the FANN and ANN models the results confirm the usefulness of the factors that were extracted from a large set of related. On the other hand, as far as estimation is concerned the nonlinear FANN model is more suitable to capture nonlinearity and structural breaks compared to linear models. The Diebold-Mariano test results confirm the superiority of the FANN model forecasts performance over the AR benchmark model and the ANN model forecasts.