Multi-model-driven prediction of magnetic phase transitions and magnetocaloric effects in NiMnFeCoBP high-entropy amorphous alloys

Accurate prediction of magnetic phase-transitions is essential for the applicability of the magnetocaloric effect. Despite the demonstrable efficacy of machine learning in addressing such issues, existing strategies remain constrained to specific material categories, exhibiting limited generalizability across diverse systems. Herein, we propose a multi-model ensemble framework that overcomes the limitations of the conventional single-model paradigm in NiMnFeCoBP high-entropy-amorphous-alloys. The integration of complementary methodologies has yielded a 9%-13% increase in prediction accuracy when utilizing an ensemble model compared with single models. This adaptive strategy effectively resolves the accuracy-generality trade-off dilemma in materials informatics by leveraging the collective strengths of multiple predictive models.