Newtonian mechanics based hybrid machine learning method of characterizing energy band gap of doped zno semiconductor

Abstract

Listen

Translate article

This present work utilizes the distortion due to the incorporation of dopants into crystal lattice structure of zinc oxide (ZnO) semiconductor to model the associated energy band gap. The proposed model hybridizes machine learning support vector regression model (MLSVRM) and gravitational search algorithm (GSA) that is based on Newtonian mechanical principle of motion. The performance of the developed hybrid gravitational search (GS) based MLSVRM is compared with stepwise regression (ST) based model and the existing ordinary support vector regression computational intelligence (SVRCI) model in the literature. The developed hybrid GS-MLSVRM performs better than ST-based model and the existing SVRCI model in the literature by 83.74% and 42.59%, respectively using mean absolute percentage deviation (MAPD) as a performance measuring parameter. The developed hybrid GS-MLSVRM further shows superior performance over the two compared models using other performance measuring parameters such as root mean square error (RMSE), mean absolute error (MAE) and correlation coefficient (CC). The superior performance of the developed hybrid model can be attributed to the power of hybridization and intrinsic ability to approximate non-linear relationship between lattice distortion and energy band gap of the doped semiconductor. The precision of the developed hybrid model would ultimately hasten ZnO band gap characterization without experimental stress.