Prediction of elastic properties of the crystal structure of Heusler alloys using machine learning

This study focuses on applying machine learning to predict the mechanical properties of Heusler alloys. The target parameters are the bulk modulus ($B$) and Poisson’s ratio ($\nu$). Models were built based on various algorithms, including linear methods, tree-based ensembles, and neural networks. The best performance was achieved using gradient boosting: $R^2$ = 96.4 % and RMSE = 14.9 GPa for $B$; $R^2$ = 65.9 % and RMSE = 0.035 for $\nu$. The models were validated on an independent dataset of 965 Heusler alloys, including all-$d$ structures. The results confirm the applicability of the proposed approach for the preliminary screening of mechanically stable candidate materials.