cugbczg/Machine-Learning-to-train-models-tracing-fluids-in-LIPs: v.4.0.0

Random forest (RF), deep neural network (DNN) and support vector machines (SVM) are employed trace the role of fluids in the generation of the early Permian Tarim large igneous province (LIP) in northwestern China. Training data were extracted from the GEOROC (http://georoc.mpch-mainz.gwdg.de/georoc/) and PetDB (https://search.earthchem.org/) databases involving global island-arc basalts (IAB), mid-oceanic ridge basalts (MORB) and OIB to establish the predictive models. Firstly, train models based on RF model using 10194 basalt samples, comprising 3277 IAB samples, 5920 OIB samples, and 997 MORB samples. Secondly, basaltic samples were tested to yield the prediction results.

BASALT_part1-8, earthchem_download_76508--major element and earthchem_download_84755--trace element are the primary data collected before data cleaning. Supplementary Dataset S1. Compilation of geochemical elements data for IAB, OIB and MORB from GEOROC and PetDB databases after data cleaning. Supplementary Dataset S2: Compilation of geochemical elements data for Tarim LIP and CAOB before data cleaning. Supplementary Dataset S3: Compilation of geochemical elements data for Tarim LIP and CAOB after data cleaning.