Garnet major-element composition as an indicator of host-rock type: a machine learning approach using the random forest classifier

Abstract The major-element chemical composition of garnet provides valuable petrogenetic information, particularly in metamorphic rocks. When facing detrital garnet, information about the bulk-rock and mineral paragenesis initial garnet-bearing host-rock is absent. This prevents application thermo-barometric techniques calls for quantitative empirical approaches. Here we present a discrimination scheme that based on random forest machine-learning algorithm trained large dataset 13,615 analyses covers wide variety lithologies. Considering out-of-bag error, correctly predicts original (i) > 95% concerning setting, either mantle, metamorphic, igneous, or metasomatic; (ii) 84% facies, blueschist/greenschist, amphibolite, granulite, eclogite/ultrahigh-pressure; (iii) 93% bulk composition, intermediate–felsic/metasedimentary, mafic, ultramafic, alkaline, calc–silicate. coverage potential host rocks, detailed prediction classes, high rates, successfully tested real-case applications demonstrate introduced overcomes many issues related to previous schemes. highlights transferring applied strategy broad range minerals beyond garnet. For easy quick usage, freely accessible web app provided guides user five steps from results including data visualization.