A Benchmark Dataset for Machine Learning–Based Prediction of Hydrogen Evolution Efficiency from GLFO-Catalyzed Sucrose Photocatalysis

This study introduces a novel experimental dataset designed to support the development and evaluation of artificial intelligence models for photocatalytic hydrogen production. The dataset was generated from a series of controlled laboratory experiments utilizing a graphene-supported LaFeO₃ (GLFO) catalyst and sucrose solution as the organic sacrificial agent. Key process variables—including solution pH, catalyst concentration, and initial hydrogen peroxide concentration (HPC)—were systematically varied using a Box–Behnken design to investigate their influence on hydrogen evolution. A total of 1370 data points were collected, each representing a unique combination of input parameters and the corresponding hydrogen yield (mmol gcat⁻¹). The dataset provides high-resolution insight into nonlinear interactions in the hydrogen generation process and is intended as a benchmark resource for reproducible research in machine learning-driven photocatalysis. Initial analysis confirms the dataset’s suitability for training predictive models and evaluating optimization strategies. By releasing this dataset, we aim to advance reproducible research and accelerate the application of AI in clean energy technologies, particularly in the field of photocatalytic hydrogen production.