Dataset and Reproducibility Materials for Uncertainty-Aware Surrogate-Assisted Learning in Deep Eutectic Solvent Battery Recycling

This repository contains the curated dataset and reproducibility materials associated with the study “Uncertainty-Aware Surrogate-Assisted Learning for Metal Recovery Prediction in Deep Eutectic Solvent Battery Recycling.” The original literature compilation contained 622 experimental records collected from 39 independent peer-reviewed studies. Following data validation and removal of seven duplicate observations, 615 unique records were retained and used for all machine-learning experiments. The dataset covers lithium, cobalt, nickel, and manganese recovery using deep eutectic solvent systems. The repository includes the curated modelling dataset, source-aware cross-validation assignments, experiment configuration files, software-version information, fold-level predictions and metrics, Gaussian Process Regression results, uncertainty-aware augmentation and ablation results, model-interpretation outputs, analysis notebooks, and supporting manuscript tables and figures. The publicly released curated dataset corresponds to the exact 615 observations used in the reported experiments. The dataset is literature-derived. Original experimental observations should be traced to and cited through the corresponding source publications identified in the dataset documentation.