Machine Learning Model to Predict the Formation Energy of Copper-based Ternary Alloys

Formation energy plays a crucial role in material development, serving as a key metric for understanding material stability and behavior. In this study, machine learning algorithms, namely Random Forest Regressor (RFR) and Gradient Boosting Regressor (GBR), were employed to predict the formation energy of copper-based ternary alloys. The models were implemented using the Scikit-Learn library within the Anaconda distribution. A composition-based featurizer, Magpie elemental properties, from the Matminer toolkit, were utilized to represent the alloy's features. The results demonstrate that the composition-based featurizer effectively captures the relationship between alloy composition and formation energy. Among the models, GBR outperformed RFR, explaining 94% of the variance in formation energy using only five features, compared to 92.5% explained by RFR, which required ten features. These findings highlight the efficiency and accuracy of GBR in predicting formation energy with fewer input features. This work underscores the potential of machine learning models, particularly the GBR, as powerful tools for accelerating material discovery and design. By enabling reliable and efficient predictions, these models provide a pathway to streamline material development processes.