# Machine Learning Phase Classification of Thermoelectric Materials

## TL;DR

This paper uses machine learning to quickly classify phases of thermoelectric materials, helping speed up their development.

## Contribution

A Support Vector Machine model is introduced for efficient phase classification of thermoelectric alloys.

## Key findings

- The SVM model achieves prediction accuracies between 77% and 92% for thermoelectric alloy phases.
- Cross-validation confirms the model's robustness in differentiating various TE phases.

## Abstract

In this study, we employ a Support Vector Machine (SVM) model to efficiently classify the phases of thermoelectric (TE) alloys. While ab initio calculations and experiments have explored the phases of functional TE materials, the large variety of alloys makes these explorations time-consuming and expensive. Therefore, there is a critical need for time-efficient methods to accelerate the discovery and development of new TE materials. Recently, machine learning (ML) classification models have been applied to predict material phases, including those of multi-principal element alloys. Using an SVM to classify phases of TE alloys, our results demonstrate that the model achieves prediction accuracies ranging from 77% to 92%. Additionally, cross-validation across various TE phases is performed to demonstrate the model’s robustness in phase differentiation. This work offers a time-efficient computational approach to distinguish TE material phases, offering valuable insights that can aid in the evaluation and design of high-performance thermoelectric materials.