Machine Learning Framework and Dataset for Titanium Alloy Thermal Conductivity Prediction

This repository contains the datasets and Python Jupyter notebooks utilized to develop a physics-informed machine learning framework for predicting the thermal conductivity of titanium alloys. The repository includes: Training Dataset: 309 unique titanium alloy samples extracted from the MatWeb database, featuring engineered thermal processing descriptors. Validation Dataset: 47 independent samples curated from published literature for strict applicability-domain testing. Primary Model (With Ti.ipynb): The complete Nested GroupKFold cross-validation and CatBoost modeling pipeline. Ablation Model (WITHOUT Ti.ipynb): The sensitivity analysis pipeline demonstrating the predictive degradation when the base titanium matrix feature is removed.