基于机器学习的榍石微量元素特征判别岩浆岩类型OA北大核心

Titanite,as an accessory mineral in igneous rocks,provides a record of the evolution of geochemical environments during magma differentiation,and its trace element characteristics hold significant potential in rock type identification.Although binary discrimination plots have been applied in traditional classifications of igneous rocks based on titanite trace element content,their limitations in classifying complex rock types have become increasingly evident.In recent years,with the accumulation of geological data and the development of machine-learning technologies,data-driven geological research methods have been widely applied.This study utilized publicly available titanite trace element data and employed genetic programming symbolic regression to construct a mathematical expression for discriminating igneous rock type.By applying four machine-learning algorithms(support vector machine,k-nearest neighbors,XGBoost,and random forest),a classification model for igneous rock type was constructed,which accurately distinguishes ultramafic,mafic,intermediate,and felsic rocks.SHAP analysis revealed significant associations between titanite trace element compositions and igneous rock types,indicating that titanite geochemistry can serve as an effective indicator for lithological classification.