Steel price volatility forecasting; application of the artificial neural network approach and GARCH family models

GARCH family models are the most widely-used methods for forecasting price volatility. Given that this approach usually has extremely high forecast errors, continuous studies have been conducted to improve forecast models using different techniques. In the present manuscript, we expanded the fields of expert systems, forecast, and modeling using an artificial neural network (ANN) and Generalized Autoregressive Conditional Heteroskedasticity (GARCH) method that created an ANN-GARCH model. The hybrid ANN-GARCH model was used to forecast steel price volatility, and its accuracy was evaluated based on mean absolute error (MAE) and mean square error (MSE) evaluation criteria. The results indicated a general improvement in forecasting using ANN-GARCH compared to the GARCH method alone. The results were realized using copper price returns, the dollar index, gold price returns, and oil price returns as inputs. We also discussed the research implications for this field in addition to practical applications. The research results indicated better performance of the hybrid ANN/GARCH/N model than other models. Furthermore, the neural-network-based hybrid models could better forecast prices than other time series models.

Keywords:

Steel price volatility , Forecast , The Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model , The exponential general autoregressive conditional Heteroskedasticity (EGARCH) model , Artificial neural network (ANN) , Hybrid model