| Abstract Scope |
Artificial intelligence (AI) studies to predict or suggest efficient thermoelectric materials have become increasingly important. We developed a machine learning pipeline trained with multivariable inputs on a massive public dataset to predict total thermal conductivity using four test sets: three publicly available datasets and one dataset created using all our own previous results. Due to the inclusion of these massive datasets, our model presents a promising possibility to further expand the understanding of the selection of features with various thermoelectric materials. Additionally, with the aid of feature selection and importance analysis, useful chemical features were chosen that ultimately led to higher accuracy in the test sets. With this contribution, we present the extension onto the thermoelectric figure-of-merit, which brings us into the exciting situation to predict the performance simply based on the chemical formula. |