Challenging handheld NIR spectrometers with moisture analysis in plant matrices: Performance of PLSR vs. GPR vs. ANN modelling

The global demand for natural products grows rapidly, intensifying the request development of high-throughput, fast, non-invasive tools quality control applicable on-site. Moisture content is one most important parameters products. It determines their market suitability, stability and shelf life should preferably be constantly monitored. Miniaturized near-infrared (NIR) spectroscopy a powerful method on-site analysis, potentially fulfilling this requirement. Here, feasibility study applicability analytical performance three miniaturized NIR spectrometers two benchtop instruments was evaluated in that scenario. case involved 192 dried plant extracts composed five different plants harvested countries at various times within years. reference analysis by Karl Fischer titration determined water sample set between 1.36% 6.47%. For spectroscopic half samples were laced with drying agent to comply industry standard. calibration models on basis root-mean square error prediction (RMSEP) an independent test set. Partial least squares regression (PLSR), Gaussian process (GPR) artificial neural network (ANN) constructed spectral sets from each instrument. GPR ANN performed superior all measured handheld native ones analyzed instruments. Moreover, accuracy penalty when analyzing lower as well. With or calibration, offered level Therefore, application can used no vs. laboratory-based analysis.