Application of XGBoost using data from the oxygen reduction reaction of carbonaceous materials for H2O2 electrosynthesis

Carbon-based materials, such as carbon black, have emerged as prominent candidates for the cathodic electrosynthesis of H 2 O 2 due to their low cost, durability, and ease of chemical modification. While recent advances in catalyst design and parameter optimization have improved the electroasynthesis efficiency, the large number of interdependent features introduces complexity and unpredictability in the development of the system. In this context, machine learning (ML) algorithms have demonstrated great potential when used in interpreting complex datasets and deriving meaningful insights in diverse areas, including electrochemistry. While the selectivity and efficiency of metal-free carbon-based catalysts are strongly influenced by their surface functional groups, having a precise control over these functional groups remains a challenging task, and this affects the interpretation of experimental results. In this study, the XGBoost algorithm was used to analyze literature data on experimental conditions for H 2 O 2 electrosynthesis, physicochemical characterizations, and the application of materials and their influence on H 2 O 2 production efficiency. XPS analysis, Raman spectroscopy, BET surface area analysis, contact angle measurements, and electrical property analyses were used to characterize the electrocatalysts. The developed model demonstrated strong generalization capabilities, even when tested on external datasets not included in the training program. Furthermore, hyperparameter optimization analysis was used to refine the database, where the most relevant catalyst features were identified. This analysis enabled the creation of a more efficient model with reduced computational demands. Finally, thorough discussions were put forth regarding the challenges involving the construction of a robust database from fragmented experimental literature. This work provides useful contributions and insights into the application of ML in H 2 O 2 electrosynthesis from experimental data, the construction of a structured database of carbonaceous catalysts, and the systematic interpretation of modeling results, paving the way for the rational design of new materials. • Machine learning model using XGBoost predicts H 2 O 2 electrosynthesis efficiency. • Analysis integrates multifaceted characterization data of carbon-based catalysts. • Key catalyst features identified through SHAP value analysis. • The model demonstrates strong generalization on an external validation dataset. • Framework addresses challenges in constructing robust databases from literature.