Modeling of semiconductors refractive indices using hybrid chemometric model

Abstract

A support vector regression (SVR)-based model and its hybrid (HSVR), both optimized with gravitational search algorithm (GSA), for accurate estimation of refractive indices of semiconductors using their energy gaps as descriptors are presented. The proposed GSA-HSVR model demonstrates a better predictive and generalization ability than ordinary GSA-SVR model. The performances of the proposed models are compared with the existing Moss and Ravindra's models and a better agreement with the experimental values were observed coupled with lowest mean absolute error of GSA-HSVR model. Considerable high coefficient of correlation and very small root mean square error also characterize GSA-HSVR model. The proposed GSA-HSVR model proves its identity and effectiveness compared to existing predictive models, in terms of accuracy, using simply accessible descriptor. It also reduces the estimation challenges accompanying determination of refractive indices of semiconductors.