Research on Property Prediction of Materials Based on Machine Learning

Abstract

In this paper, three feature selection methods and three machine learning regression models are used to select the best feature subset from the feature set to predict compound energy performance. By comparing the matching performance of different feature selection methods and machine learning models, the experimental results show that Ant Colony Optimization is used for feature selection and the Support Vector Regression model is used for the best prediction effect. The research in this paper can provide references for the prediction of new material properties in the future.