A Practical Approach for Tectonic Discrimination of Basalts Using Geochemical Data Through Machine Learning

Identifying the tectonic setting of formation of rocks is an essential component in the field of geosciences. The conventional approach is to employ standard tectonic discrimination diagrams based on elemental correlations and ratios, which sometimes are plagued with uncertainties and limitations. The application of machine learning algorithms based on big data can effectively overcome these problems. In this study, three machine learning algorithms, namely Support Vector Machine, Random Forest, and XGBoost, were employed to classify the various types of basalts from diverse settings such as intraplate basalts, island arc basalts, ocean island basalts, mid-ocean ridge basalts, back-arc basin basalts, oceanic flood basalts, and continental flood basalts into seven tectonic environments. For the altered basalts and fresh basalt, we use 22 relatively immobile elements (TiO2, P2O5, Nb, Ta, Zr, Hf, Y, La, Ce, Pr, Nd, Sm, Eu, Gd, Ho, Er, Yb, Lu, Dy, Tb, Cr, Ni) and 35 major plus trace elements to build discrimination models for seven types of tectonic settings of basalt, respectively. The results indicate that XGBoost demonstrates the best performance in discriminating basalts into seven tectonic settings, achieving an accuracy of 85% and 89% respectively. Compared to previous models, our new method presented in this study is expected to have better practical applications.