Application of artificial neural networks for the categorization of mineral resources in a copper deposit in Peru

Purpose The purpose of this paper is to categorize mineral resources in a copper deposit in Peru using an artificial neural network (ANN). Design/methodology/approach In this work, the categorization process integrates a two-step machine learning framework. First, the K-Prototypes algorithm is applied to cluster blocks based on geological and spatial characteristics. Subsequently, a multilayer perceptron ANN refines the classification by smoothing categorical boundaries, ensuring spatial coherence. The model was trained using 318443 blocks, with its performance evaluated through accuracy, recall and F1-score metrics. Findings The ANN achieved an overall accuracy of 93%, demonstrating superior classification reliability. The Measured category exhibited the highest precision (0.96) and F1-score (0.97), while the Indicated and Inferred categories achieved balanced F1-scores of 0.90, reflecting minor classification overlap. The estimated total tonnage was 5859.35 Mt, distributed as 1395.99 Mt (Measured), 2208.72 Mt (Indicated) and 2254.64 Mt (Inferred). The corresponding fine copper content was 5.40 Mt, 6.56 Mt and 6.29 Mt, respectively, with average grades of 0.43%, 0.33% and 0.31% Cu. The ANN reduced classification boundary discontinuities, enhanced geological consistency. Originality/value This study introduces a machine learning approach that integrates clustering and deep learning to improve resource classification accuracy and spatial consistency, offering a more reproducible and scalable alternative to traditional methods.