Development of Predictive Regression Models in the Copper Mining Industry Using Artificial Neural Networks (ANN)

Abstract

The present research focused on the development of artificial intelligence (AI) tools within the mining sector, specifically through predictive regression models, with an emphasis on the copper industry. This study adopted a quantitative and exploratory approach, enabling the development of models using various features of artificial neural networks (ANN) through Machine Learning algorithms, based on historical data in both cases. The main objective of this study was to develop predictive regression models in the copper mining industry using artificial neural networks (ANN). In this context, forecasts were made for the future price of copper and the capital cost of open-pit copper mines. The results indicate that it was possible to estimate, with an acceptable degree of accuracy, both the future price of copper and the mining capital cost for open-pit copper mines. In the first case, a model was developed using the TensorFlow-based Deep Learning ANN technique for the "REDPC_17" test. A 7-day forecast was achieved with an acceptable margin for the three networks—LSTM, GRU, and CNN—resulting in an MSE of 2.6621e-04, 5.8286e-04, and 3.0642e-04, respectively. This led to an average percentage error of 7.9342% for the CNN network, which proved to be the best performer in the test. In the second case, a model was developed using the Scikit-Learn MLPRegressor ANN technique for the "REDES_20" test. This model achieved the best performance in estimating CAPEX, with a determination coefficient (R2) of 0.00960 and an MSE of 0.8674.