The nature of statistical learning theory
The nature of statistical learning theory
A tutorial on support vector regression
Statistics and Computing
Combination of support vector machines using genetic programming
International Journal of Hybrid Intelligent Systems
| Hi-index | 0.00 |
Due to problems arising from lattice mismatch in thin film growth in semiconductor manufacturing industry, researchers have put sizeable efforts to reduce error in the prediction of lattice constant (LC) of cubic perovskites. However, linear prediction models were developed using linear regression techniques, which may not be able to find precisely the underlying nonlinearity in correlating LC to atomic parameters of perovskites. This causes a reduction in the prediction accuracy of linear model. To address this problem, in this work, support vector regression (SVR) technique is proposed to design and develop LC prediction model for cubic perovskites. To investigate the generalization of SVR model, we collected data for new cubic and pseudocubic compounds from the current literature of material science. Our analysis shows an improved prediction performance of SVR models than existing linear models. The proposed SVR model demonstrates lower overall error values of 0.286, 0.528, and 0.615 for training, validation, and newly collected compounds, respectively.