Machine learning-aided design of LaCe(Fe,Mn,Si)₁₃H-type magnetocaloric materials for room-temperature applications

La(Fe,Si)₁₃ magnetocaloric materials have attracted extensive attention in recent years due to their excellent magnetocaloric properties and low price. In this study, we established and compared several machine models and used a support vector machine for predicting the influences of different element contents on the Curie temperature. Based on the results predicted by the machine learning model, a series of fully hydrogenated materials (La_1.1-xCe_xFe_13-y-zMn_ySi_zH_F, y≈0.3–0.5x, z=1.1, 1.3, 1.5) with Curie temperatures between 290 and 310 K were designed. After material screening, a material with an element composition of La_0.66Ce_0.44Fe_11.4Mn_0.1Si_1.5H_F was proven to have good room temperature magnetocaloric properties, with a Curie temperature of 301 K, magnetic entropy change of 11.3 J/kgK under a 0–2 T magnetic field, and a relative cooling power of 119.3 J/kg, which performs well in the existing La(Fe,Si)₁₃-type materials that can be applied near room temperature. This work not only deepens researchers' understanding of machine learning-aided design of materials with specific application environment restrictions but also greatly accelerates the design of near-room temperature La(Fe,Si)₁₃ magnetocaloric materials.