Gaining Insight from Semi-Variograms into Machine Learning Performance of Rock Domains at a Copper Mine

Machine learning (ML) is increasingly being leveraged by the mining industry to understand how rock properties vary at a mine site. In previously published work, type, granodiorite, was predicted with high accuracy random forest (RF) ML method Erdenet copper in Mongolia. As result of optimistic results (86% overall success rate), this paper extended research determine if would be successful modeling domains. Rock domains are groups rocks that occur together. There were two additional goals. One variograms could predict or help methods perform on data. The second 2D well given disseminated nature deposit. methods, multilayer perceptron (MLP), k-nearest neighborhood (KNN) and RF, applied model six domains, D0–D5, 3D. Modeling performance poor 2D. Prediction 3D for D1 (92–94%), D2 (94–96%) D4 (85–98%). Note together constituted about 80% samples. Conclusions drawn based since poor. appeared depend factors. It better domain when not minuscule proportion sample. also whose indicator semi-variogram (ISV) range high. For example, though only contributed 15% samples, MLP did as KNN RF performing best. hyperparameters suggested best small number samples used make prediction. summary conclusion most important D2, using ML. semi-variograms can provide insight into performance.