using machine learning for prediction of viscosity of CO2-N2 mixture

This repository provides a minimal, self-contained reproducibility package for the machine-learning models developed to predict the viscosity of CO₂–N₂ gas mixtures. It includes the final curated dataset (N = 2912) as well as the main data (Viscosity_Cleaned with a data dictionary, fixed-seed Python scripts implementing ensemble regression models (Gradient Boosting, Random Forest, XGBoost, LightGBM, CatBoost, and Stacking), and MATLAB scripts implementing multi-algorithm artificial neural networks (LM, SCG, and Bayesian Regularization). All stochastic components, including data partitioning, model training, and bootstrap-based uncertainty analysis, are executed using fixed random seeds to ensure deterministic reproducibility. The package reproduces all tables and figures reported in the manuscript, including uncertainty-aware Accumulated Local Effects (ALE) analyses for temperature, pressure, and CO₂ composition, and documents the computational environment and software versions used.