Hardness Prediction of MoNbTaW Alloy Films Based on Machine Learning and Interpretability Analysis

Machine learning (ML) offers a powerful paradigm for accelerating performance prediction of high-entropy alloys (HEAs). The present study proposed an ML framework based on the ridge regression algorithm for predicting the hardness of MoNbTaW HEA films. By comparing various feature-screening strategies, an optimized feature set comprising three features, namely <i>δG</i>, <i>Λ</i>, and <i>Ω</i>, was selected from 20 candidate physical features. The model based on this feature set exhibited strong predictive performance. In 10-fold cross-validation, <i>R</i>² was 0.86, <i>RMSE</i> was 0.41 GPa and MAE was 0.31 GPa. On the reserved validation set, <i>R</i>² was 0.88, <i>RMSE</i> was 0.37 GPa, and MAE was 0.31 GPa. The model further revealed the influence trends of constituent elements and key features on hardness. By using ML to mine useful information from a dataset of HEA film samples prepared via magnetron sputtering, this work provides an approach for rapid and cost-effective design of HEAs.