Abstract

In the copper mining industry, Data Science (DS) techniques and Machine Learning (ML) methods are contributing to improve the prediction of results in industrial processes. In this paper, an experience of applying both DS techniques and a ML algorithm, using historical data from the flotation process is described. These data were collected using a prototype of flotation equipment developed at the Universidad Catolica del Norte, in Antofagasta, Chile. To achieve the result an Extraction, Transformation and Load (ETL) process was made. Also, for both, improving the understanding of domain dynamics and selecting the most relevant predictive variables in the flotation process, a Random Forest (FR) model was developed. The combination of these previous results made it possible to generate recommendations on the management of predictor variables to improve copper recovery in the context of the flotation equipment prototype. In this document, the methodological details are presented, and the process used to obtain the aforementioned results is described. As progress was made through 2 iterations, the quality of the results obtained with the predictive model, generated by RF, was improving. At the end of the process, an accuracy of 94,44% was achieved, with an accuracy in each of the classes greater than 90%. These results demonstrate the effectiveness and outstanding performance of the predictive model. These values are highly competitive when compared to those obtained in other similar studies in the context of Industry 4.0.