Improved machine learning framework for prediction of phases and crystal structures of high entropy alloys

High-entropy alloys (HEAs) are gaining popularity because of their remarkable properties controlled by phases and crystal structures. In addition to that, in the field of material informatics, machine learning (ML) techniques have gained considerable attention in predicting phases and crystal structures of HEAs. In this study, a novel ML-based methodology has been proposed to predict different phase stages and crystal structures. To this end, 1345 data samples were used to train the ML model to predict the phases of HEAs. Within the dataset, 705 data were utilized to predict the crystal structures with the help of thermodynamics and electronic configuration as input features. The important features were selected using the Pearson correlation coefficient matrix, followed by using of five distinct boosting algorithms to predict phases and crystal structures. Among all these algorithms, XGBoost recorded the highest detection accuracy of 94.05 % for phases and LightGBM yielded the highest detection accuracy of 90.07 % for crystal structure. Various hyperparameter tuning was conducted to find the optimum performance of the boosting classifiers. A comprehensive comparison was performed between the ML models and some from published papers in reputed journals. From the comparison, it was evident that the proposed methodology showed its superiority in terms of phase and crystal structure detection of HEAs.