Prediction of vitamin C using FTIR-ATR terahertz spectroscopy combined with interval partial least squares (iPLS) regression

In this study FTIR-ATR terahertz (THz) spectroscopy combined with interval PLS (iPLS) regression was used to measure concentration of vitamin C in aqueous solution. iPLS regression was used to select the efficient spectral regions and variables to develop calibration model. The performance of the iPLS model was then compared to that of full-spectrum PLS model. The result obtained by iPLS model with 5 PLS factors was superior than that of full-spectrum PLS model with 11 PLS factors when 7 spectral regions and 70 variables were selected. Prediction performance of vitamin C can be improved by using iPLS model with ratio prediction to deviation (RPD) value of 4.570. This work demonstrated that concentration of vitamin C in aqueous solution can be predicted by FTIR-ATR THz spectroscopy method, and iPLS regression method revealed its superiority in model calibration.