Machine learning based prediction of specific heat capacity for half-Heusler compounds

Half-Heusler alloys are among the most emerging families due to their different properties in topological insulators, superconductors, and magnetic behavior, which are directly applicable to developing low-cost and high-power spintronics devices. This study investigates the predictive performance of a stacked model for estimating the lattice parameters and specific heat capacity of 438 half-Heusler alloys with 28 columns in different properties. The stacked model, which incorporates gradient boosting and random forest as baseline models, was meticulously tuned for parameter optimization. Our calculated results demonstrate the robustness of our model, as evidenced by the high R-squared scores that indicate remarkable accuracy and consistency in predicting lattice parameters and specific heat capacity. The model also shows strong correlation coefficients, underscoring its reliability and precision. Comparative analysis reveals the superiority of the stacked model over alternative approaches, positioning it as the preferred model for both properties. This research highlights the stacked model’s efficacy in material property prediction, offering valuable insights for materials science research and development at a very low cost.