Comparison analysis between PLS and NN in noninvasive blood glucose concentration prediction

A series pair data of NIR spectral and measured BGL are collected for an OGTT experiment from a healthy volunteer. The collected data are then calibrated by using partial least squares (PLS) regression and feed-forward back-propagation neural network (NN). A comparative analysis between both calibration models is analysed. From the PLS and NN calibration models, root mean square error prediction of 0.5282 mmol/L and 0.2952 mmol/L, respectively, were achieved. The correlation factor of 0.9247 and 0.9863 were obtained from PLS and NN calibration models respectively.