Non-linear Neural Models to Predict HRC Steel Price in Spain

Abstract

Steel is a raw material widely used in industry due to its advantages over other alternatives, such as cost, fast and environmentally friendly recycling, ease of use, high strength, different finishes and qualities. Forecasting steel prices has been an important and challenging task that has traditionally been tackled with econometric, stochastic-Gaussian and time series techniques. Advancing from previous work on this open challenge, in the present paper some Artificial Neural Networks are applied for the first time to forecast the price of hot rolled steel in Spain. More precisely, some non-linear neural networks are applied to several different input time series. The target of this research is twofold; on the one hand, identify which of the neural models outperforms the other ones when predicting steel prices and, on the other hand to validate different data series for such prediction. The main outcomes of this research, after validating the neural models on real data from last 7 years, greatly contribute to this field as novel and relevant conclusions are obtained.