Accelerated search for ABO3-type the electronic contribution of polycrystalline dielectric constants by machine learning
| Autor: | Xinyu Lin, Changjiao Li, Hanxing Liu, Hua Hao, Guanghui Zhao |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Work (thermodynamics)
General Computer Science Mean squared error Band gap General Physics and Astronomy 02 engineering and technology Dielectric Type (model theory) 010402 general chemistry Machine learning computer.software_genre 01 natural sciences Feature (machine learning) General Materials Science business.industry General Chemistry 021001 nanoscience & nanotechnology Abstract machine 0104 chemical sciences Computational Mathematics Mechanics of Materials Gradient boosting Artificial intelligence 0210 nano-technology business computer |
| Zdroj: | Computational Materials Science. 193:110404 |
| ISSN: | 0927-0256 |
| DOI: | 10.1016/j.commatsci.2021.110404 |
| Popis: | Machine learning method has made rapid progress in current material computing science and has become a popular learning tool. In this work, machine learning models are employed to predict the electronic contribution of the polycrystalline dielectric constant. The value of the electronic contribution of polycrystalline dielectric constant is accurately predicted via the Gradient boosting regression (GBR) model by using only 42 features of simply structural and elemental information. The predicted performance of the GBR model reaches R2 of 0.887, MAE of 0.492, MAPE of 0.081, and RMSE of 0.824, which far exceeds other models. With the analysis of the feature importance, the band gap shows a large correlation with the electronic contribution of polycrystalline dielectric constant from the data perspective. This method proposes a machine learning model for the polycrystalline dielectric constant creatively, bypasses the time-consuming first principles calculations and the complex analysis of materials internal connections, which provides a certain idea and method for the application of machine learning in dielectric materials and accelerates the research of dielectric materials. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect