Machine Learning Based Structural Design and Performance Prediction of Advanced Materials

This paper explores the core methods of machine learning in material modelling, structure-property correlation analysis and optimal design, including data pre-processing, feature engineering, algorithm selection and model training strategies. Combining deep learning, graph neural network (GNN) and gradient boosting decision tree (GBDT) methods, accurate prediction of material properties can be achieved and material structures can be optimised by Bayesian optimisation, genetic algorithms and other intelligent search strategies. It is shown that the machine learning-based material design method can significantly improve the efficiency of material screening and accelerate the development process of new materials.