Quantitative evaluation of glucose spectra from NIR spectroscopy measurements using PLS regression analysis

The quantitative evaluations were carried out in NIR spectroscopy that was implemented for monitoring and predicting the concentration of glucose samples. The collected absorbance data was preprocessed in developing PLS model before calibration using Savitzky-Golay filter. The spectrum was corrected by subtracting the offset of the regression to the absorption value and dividing this difference by the slope using Leave One Out Cross Validation (LOOCV) of the training set to determine the optimum number of PLS components. The Samples of glucose solution consist of 21 different molarity from 3000 to 5000 mg/dl with the interval of 100 mg/dl in step. Results obtained shown the linear dependency of the reference and predicted glucose concentration, with RMSECV and R2CV value are 104.92 mg/dl and 0.9728, respectively. The RMSECV shown the lowest error present and R2CV were close to one, indicates that the PLS model suited to accurately predict the variability glucose concentration.