1. Prediction of superconducting properties of materials based on machine learning models
- Author
-
Yongquan Jiang, Hu Jie, Yang Yan, Yu Siqi, and Zuo Houchen
- Abstract
Superconducting materials have extremely high application value in our life. At present, the discovery of new superconducting materials relies on the experience of experts through a large number of "trial and error" experiments. Obtaining the properties of superconductors also requires a large number of experiments. In this paper, we propose to use the XGBoost model to identify superconductors, reaching 0.986 in accuracy. We apply the deep forest algorithm to predict the critical temperature of superconductors and the coefficient of determination reaches 0.944. We propose to apply the same algorithm to predict the forbidden band width of materials and the coefficient of determination reaches 0.917. A new sub-network structure is constructed to predict the Fermi level of materials and the coefficient of determination reaches 0.984. All of these algorithms have state-of-the-art performance. Finally, the model is tested with a publicly available dataset to identify 50 candidate superconducting materials with a critical temperature greater than 90K.
- Published
- 2022
- Full Text
- View/download PDF