A physics-driven machine learning framework for predicting XRD patterns, magnetization, and magnetocaloric response...

Rare-earth-doped double perovskite oxides have recently emerged as highly promising materials for magnetic refrigeration, owing to their remarkable structural complexity and strongly correlated magnetic behaviors. Accurately predicting their physical properties from experimental data remains a key challenge for accelerating materials discovery and costly experimental efforts. In this work, we introduce a comprehensive machine learning framework for the prediction of the structural, magnetic, and magnetocaloric properties of the La₁.₉₇₅Ce₀.₀₁₂₅Er₀.₀₁₂₅NiMnO₆ (LCENMO) double perovskite compound. Five supervised machine learning algorithms were systematically investigated, namely Decision Tree (DT), Random Forest (RF), Gradient Boosting (GB), Extra Trees (ET), and Multi-Layer Perceptron (MLP), and trained on experimentally derived datasets encompassing X-ray diffraction (XRD) patterns, temperature-dependent magnetization M(T), magnetization isotherms M(T, µ ₀ H), Arrott plots, and magnetic entropy change ΔS(T, µ ₀ H). Model reliability was rigorously assessed through R ², RMSE, MAE, and 5-fold cross-validation metrics. The random forest model demonstrated outstanding predictive accuracy for XRD intensity profiles and M(T) curves, achieving R ² values of 0.996 and 0.9991 respectively, while the MLP model consistently outperformed all tree-based approaches for the prediction of magnetization isotherms, Arrott plots, and magnetocaloric entropy change, maintaining R ² > 0.99 across the full range of temperatures and applied magnetic fields. These results highlight a property-dependent complementarity between ensemble and neural network approaches, providing critical guidance for optimal algorithm selection in complex oxide systems. This work establishes a physics-guided data-driven computational framework for the LCENMO compound, paving the way toward the investigation of magnetocaloric properties in this material.