Improved Support Vector Regression Coupled with Fourier Self-Deconvolution in the Quantitative Analysis of Glucose in NIR Spectra

This paper proposes a novel calibration model based on combining the improved support vector regression coupled with the Fourier self-deconvolution for the quantitative analysis of glucose in near infrared spectra. The proposed model has been validated to predict the glucose concentration from the mixture of glucose and human serum albumin in a phosphate buffer solution. The Principal Component Regression (PCR) and the Partial Least Squares Regression (PLSR) methods are also developed under the same conditions and are compared with the proposed model. The results confirm that the proposed model yields better prediction results in comparison to the PCR and the PLSR models.