Phase Prediction in Equimolar Quaternary Medium‐Entropy Alloys via Interpretable XGBoost Modeling

Abstract Medium‐entropy alloys (MEAs) are typically composed of three to four elements, combining the advantages of both conventional alloys and high‐entropy alloys. Different phases exhibit significant differences in structure and properties; therefore, rapidly predicting phase structures from chemical compositions can greatly accelerate the discovery of novel alloys. In this study, machine learning algorithms are effectively applied to predict the phase structures of equiatomic MEAs, enabling rapid classification into single‐phase FCC/BCC structures or amorphous structures. The dataset, based on 731 structures, is analyzed using three machine learning algorithms: Random Forest Classifier (RFC), Support Vector Classifier (SVC), and Extreme Gradient Boosting (XGBoost). Hyperparameter optimization is performed using grid search with ten‐fold cross‐validation performed on the training set. The results showed that XGBoost achieved the best performance, with an accuracy of 0.939 on the test set. This indicates that XGBoost effectively identified key factors influencing phase stability and provides a reliable predictive tool for the rapid classification of equiatomic MEAs phases.