Data and Code for Machine-Learning–Based Assessment of Lithium Fertility in Himalayan–Tibetan Leucogranites

This repository contains the full dataset, code, and trained machine-learning models used to assess lithium (Li) fertility in Himalayan–Tibetan leucogranites using multivariate whole-rock geochemistry. The archive includes compiled major-oxide and trace-element datasets, compositional data analysis (CoDA–ILR) transformations, feature engineering routines, supervised and semi-supervised XGBoost classification workflows, and finalized trained models. All scripts reproduce the results presented in the associated manuscript: “Controls on Lithium Enrichment in Himalayan–Tibetan Leucogranites from Multivariate Whole-Rock Geochemistry”. The archive supports full reproducibility of data preprocessing, model training, evaluation, and batch inference.