Assessing the potential of secondary raw materials from hard coal and iron ore mining waste disposal sites using machine learning

Abstract

Machine learning is the science of how algorithms and systems improve knowledge and performance through experience. The paper discusses the possibility of applying machine learning algorithms (an AI subfield) to assess the secondary raw material potential of mining waste and tailings accumulated in hard coal and iron ore disposal sites. Applied machine learning algorithms are available for the Python language. Classification algorithms were used only (supervised machine learning). The first stage of studies included defining the category and features of the objects selected. The learning process used training data, whereas test data were applied to check the model efficiency. Analysis of the results obtained indicates that machine learning seems a promising tool to assess the secondary raw materials potential of waste accumulated in dumps and heaps, despite achieving a prediction accuracy at the level of 0.75 for hard coal objects and 0.85 for iron ore objects. It has been assumed that Gaussian NB, GaussianProcessClasifier and MLP Classifier algorithms of supervised learning show the highest prediction results. This suggests that machine learning may be used as a tool supporting the decision process, the result of which will be the economic use of waste accumulated in heaps, dumps and disposal sites.