Data and Code for: Low-Cost, High-Accuracy Reactivity Modeling: Integrating Genetic Algorithms and Machine Learning with Multilevel DFT calculations

This document provides the necessary data, software, and workflow information to reproduce the results presented in the manuscript: “Low-Cost, High-Accuracy Reactivity Modeling: Integrating Genetic Algorithms and Machine Learning with Multilevel DFT calculations”. The contents include self-contained, ready-to-run sub-directories with the necessary Python scripts, input files, and key outputs for verification.
This repository is structured for both demonstration and full reproducibility. It provides "ready-to-run" examples for direct application and foundational code templates for more complex methods (such as the GA2, GA3, and GA4 models), allowing users to reconstruct, adapt, and build upon the work.
The repository's content is organized into the following three computational levels:
Level 1: GA-ML Calibration & MLR (01_test_720LoT): Calibrates the initial hybrid framework integrating a genetic algorithm and machine learning (GA-ML) with the full 24x720 matrix and subsequently fits a Multiple Linear Regression (MLR) model.
Level 2: Reduced-Matrix GA-ML Models (02_test_576LoT): Proposes five GA-ML-based models (GA1–GA5) to optimize the selection of DFT LoTs from a reduced 24x576 matrix. This includes templates for advanced models (GA2–GA4) that use techniques such as Bayesian optimization and simultaneous regression analysis.
Level 3: Dynamic Generalization-Driven Transfer Learning (03_DGDTL): Implements a two-stage DGDTL protocol designed to enhance generalization, avoid overfitting, and find an optimal solution.