Unveiling Feature Importance in Methanol Reforming Systems through the Machine Learning

There is a common challenge in the research of methanol reforming, as the development of new catalysts requires a significant amount of time and material costs. This work leverages machine learning (ML) to explore new elements for catalysts, overcoming traditional trial-and-error methods. Utilizes machine learning to conduct in-depth feature analysis on different aspects of published methanol reforming catalyst data, including reaction conditions, elements, and descriptors. We evaluated nine ML models using four input methods to predict catalyst performance, identifying key elements and their descriptors impacting methanol reforming. The prediction reveals the importance of some previously overlooked physical properties, as well as seven potential elements that were not previously studied. And in the final preparation of the catalyst, to confirm the feasibility and accuracy of this prediction method. This novel approach sheds light on designing non-noble metal catalysts beyond conventional frameworks, providing valuable insights for methanol reforming systems.