Machine Learning Accelerates the Discovery of Efficient Oxygen Reduction Electrocatalysts High‐Entropy Intermetallic Compounds Pt(FeCoNiCuZn) ₃

ABSTRACT High‐entropy intermetallics (HEIs) exhibit great potential as efficient oxygen reduction reaction (ORR) electrocatalysts due to their great tunability, ordered crystal structures and multi‐component synergistic effects. However, the vast compositional space of HEIs poses significant challenges and difficulties to the traditional experimental and theoretical studies. Herein, taking PtM ₃ ‐type HEIs (M = Fe, Co, Ni, Cu, Zn) as a typical example, we propose a local environment‐aware feature engineering strategy integrating density functional theory (DFT) calculations and machine learning (ML) method. Specifically, *OH adsorption energies at representative surface sites were computed via DFT, and structural/compositional features were extracted as descriptors to train multiple ML models, with gradient boosting regression (GBR) exhibiting the best performance. Based on predicted *OH adsorption energies, an activity evaluation model was established by combining atomic ratio and local environment, leading to the identification of the optimal Pt ₈ Fe ₆ Co ₆ Ni ₃ Cu ₃ Zn ₆ ‐HEI. The theoretical overpotentials of ORR on the PtCu and PtZn sites of this optimal configuration were calculated to be 0.36 and 0.37 V, respectively, validating the reliability and generalization ability of the “DFT + ML” strategy. Our strategy effectively overcomes the challenge of compositional complexity in HEIs, and offers a new paradigm for the design of other high‐entropy catalytic materials, accelerating the development of advanced electrocatalysts.