Lattice thermal conductivity of material and its correlation with Debye temperature using regression models

Lattice thermal conductivity (K) of the material plays a crucial role in its variety of applications such as thermoelectric, thermal barrier coating and so on. The role of the Debye temperature (T
d
) is a significant factor in determining the lattice thermal conductivity of any materials. However, the relation between the lattice thermal conductivity and Debye temperature is very less explored using Machine Learning (ML). In this study, we first predicted lattice thermal conductivity using four machine learning regression models: Neural Network (NN), Gaussian Process Regression (GPR), Boosted Tree Regression (BTR) and Fine Tree Regression (FTR). NN and GPR is found to be the best performing model among all the used models with R
²
equal to 99.868 % and 97.333 % respectively. Further, for these two best performed models, we predict the relation between the K and T
d
in both training and testing data. The relation between training and testing exhibits similar nature in the two models NN and GPR. This work guides a pathway for searching of materials for wide applications.