Machine learning: An accelerator for the exploration and application of advanced metal-organic frameworks

Metal-organic framework (MOF) materials have the advantages of high specific surface area, large pore volume and adjustable organizational structure. It has received widespread attention in gas storage, adsorption separation, catalysis and other fields. The quantity of MOFs has shown an explosive growth trend in recent years. In addition, as a branch of artificial intelligence, the powerful adaptability, scalability, and automation of machine learning (ML) provide a powerful tool for comprehensively evaluating the application performance of MOFs in various scenarios. This makes up for the shortcomings of complex, time-consuming and safety hazards in the preparation and design of traditional porous materials. By building models using ML algorithms such as linear regression, random forests, and neural networks, it is able to predict high-performance MOFs with adsorption properties, electrical properties, catalytic properties, mechanical properties, and thermodynamics. It promotes the joint development of ML and MOFs. This review provides an overview of the general implementation methods and processes for ML assisted MOF design, including data collection, feature selection, algorithm design, and evaluation. In addition, a summary of the classic algorithms of ML and their applications in the classification and prediction for MOFs are summarized. In the end, we clarify challenges faced in current research and subsequent development directions and providing impetus for MOFs material design and development in the future.