A systematic review of machine learning approaches for phase stability and mechanical property prediction in high-entropy alloys

High entropy alloys (HEAs) have attracted significant research interest owing to their vast compositional space and diverse achievable properties. However, exploring this extensive design space and experimentally determining phase constituents and mechanical properties are time-consuming processes. Machine learning (ML) provides a data-driven approach to identifying relationships among alloy composition, phase, microstructure, and mechanical properties. This review was conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A systematic Web of Science search identified 301 records; 58 studies were selected for detailed analysis after screening. The review examined various ML models and datasets for predicting the HEA phase and mechanical properties. Most ML models have been developed to study conventional HEAs composed of Al, Co, Cu, Fe, Ni, and Mn, whereas lightweight, refractory, eutectic, and other novel HEAs remain underexplored. Prediction performance depends not only on the ML model but also on dataset size and diversity, input parameters, and target properties. Future studies should focus on developing larger, more compositionally diverse datasets, adopting transparent methods, and integrating ML models with experimental verification to improve the reliability and transferability of HEA design.