Researchers from Tokyo Institute of Technology have developed a machine learning-based clustering method for materials analysis. This new technique groups materials by considering both basic features (like composition and structure) and target properties (such as band gaps and dielectric constants). By using random forest regression, they transformed basic features into z-vectors, enabling more precise clustering. This approach, applied to over 1,000 oxides, aims to identify promising material groups for various applications and could significantly accelerate the discovery of materials with unique properties.
For more details, visit Phys.org: