Machine Learning–Based Prediction of Complex Combination Phases in High-Entropy Alloys

High-entropy alloys (HEAs) are a novel class of materials that exhibit exceptional mechanical, thermal and corrosion-resistant properties, making them highly promising for various industrial applications. The complex and variable compositions of HEAs present significant challenges in accurately predicting their phase structures, which is crucial for tailoring their properties for specific applications. This study proposes a machine learning (ML)–based approach to predict HEA phase structures using experimentally validated datasets containing chemical composition–derived features. A Boolean vector technique was employed to encode multiphase classifications, allowing the model to accurately capture complex phase relationships. Four ML algorithms consisting of support vector machine (SVM), k-nearest neighbors (KNN), random forest (RF) and neural network (NN) were utilized to develop predictive models. The models were evaluated using rigorous cross-validation (CV) and tested on unseen data samples. The results demonstrate that NN and KNN outperform other models, achieving an impressive test accuracy of 84.85%. Feature importance analysis reveals that valence election concentration and melting temperature are the most influential factors governing phase formation. This study highlights the effectiveness of ML in accelerating HEA design by significantly reducing reliance on experimental trial-and-error approaches. The predictive models can be applied in alloy design to rapidly screen compositions with desired phase structures, guiding experimental synthesis and optimizing material properties for advanced application. The findings underscore the potential of ML in computational materials science and open new avenues for integration artificial intelligence–driven approaches with conventional metallurgical techniques to enhance HEA development.