Mineral Prospectivity Mapping based on Isolation Forest and Random Forest: Implication for the Existence of Spatial Signature of Mineralization in Outliers

Known mineralized locations and randomly chosen non-mineralized are used traditionally as training samples in data-driven mineral prospectivity mapping (MPM). In this paper, we took advantage of (a) the variable importance partial dependence plot, which enable interpretation random forest (RF) modeling, (b) synthetic minority over-sampling technique, investigated efficacy outlier-based for MPM contrast to traditional practice using known positive samples. The prediction maps obtained by RF modeling based on different sets suggest bias toward MPM. proposed involve both unsupervised learning supervised learning. former aims at outlier detection, while latter uses resulting outliers investigate following: firstly, delineation prospective area or spatial signature existing system avoiding arising from secondly, coherence outliers, justifies their feasibility modeling. Analyses receiver operating curves correlations imply that derived isolation show consistent location and, thus, were effective narrowing down target areas just like