Support Vector Machines and Near Infrared Spectroscopy for Quantification of Vitamin C Content in Kiwifruit

Vitamin C is considered as an important nutrition component of fruits, especially of kiwifruit. Traditional destructive method for vitamin C measurement is very complex and fussy. This paper proposes the use of least-squares support vector machine (LS-SVM) as an alternative multivariate calibration method for the quantification of vitamin C content in “Qinmei” kiwifruit, using near infrared spectroscopy with direct measurements by diffuse reflectance in the spectral range of 800-2500 nm. The loading values for the spectral contribution of the first ten factors were used as inputs of the LS-SVM model. The best results of the LS-SVM models in this study are R2c=0.998, SEC=1.484 mg/100g for calibration and R2v=0.969, SEP=3.847 mg/100g for validation, with 10 factor used in the model. Partial least square regression (PLSR) method was also applied as a comparison. The calibration performance of the LS-SVM and PLSR models were equally well, however, the prediction performance of LS-SVM models were much better than PLSR models. It can be concluded that LS-SVM is a feasible and promising method for prediction of vitamin C content in kiwifruit from NIR spectra.