Predicting Lattice Thermal Conductivity of Transition Metals Using Machine Learning Ensemble Models

The lattice thermal conductivity (KL) of transition metals is fundamental to the applications in thermal management, high-temperature alloys, and energy-related technologies, as well as to understanding their lattice dynamics behavior. However, reliable and accurate prediction of properties remains challenging due to anharmonic lattice effects, complex structure-property relationships, and limited experimental data. This study aims to predict the KL values of transition metals from a dataset of 981 individual records that include their physical and chemical properties, with the goal of designing advanced materials. To achieve the goal, this study employed several machine learning-based models: Fine Tree (FT), Medium Tree (MT), Coarse Tree (CT), Optimizable Tree (OT), Bagged Tree (BGT), Boosted Tree (BST), and Optimizable Ensemble (OE), which were investigated on the transition metals dataset. Among the implemented models, the OE showed excellent predictive performance, achieving R² values of 88.69% in validation and 94.33% in tests for KL prediction. Furthermore, the RReliefF algorithm is employed to reveal the significance of input features to the target property (KL). Moreover, a significant correlation between KL and Grüneisen parameter was identified, and partial dependence plots were also employed to interpret feature influence and validate model predictions. These findings demonstrate the effectiveness of machine learning models in predicting thermophysical properties of transition metals.