Sensitivity for Multivariate Calibration Based on Multilayer Perceptron Artificial Neural Networks

The use of machine learning for multivariate spectroscopic data analysis in applications related to process monitoring has become very popular since non-linearities in the relationship between signal and predicted variables are commonly observed. In this regard, the use of artificial neural networks (ANN) to develop calibration models has demonstrated to be more appropriate and flexible than classical multivariate linear methods. The most frequently reported type of ANN is the so-called multilayer perceptron (MLP). Nevertheless, the latter models still lack a complete statistical characterization in terms of prediction uncertainty, which is an advantage of the parametric counterparts. In the field of analytical calibration, developments regarding the estimation of prediction errors would derive in the calculation of other analytical figures of merit (AFOMs), such as sensitivity, analytical sensitivity, and limits of detection and quantitation. In this work, equations to estimate the sensitivity in MLP-based calibrations were deduced and are here reported for the first time. The reliability of the derived sensitivity parameter was assessed through a set of simulated and experimental data. The results were also applied to a previously reported MLP fluorescence calibration methodology for the biopharmaceutical industry, yielding a value of sensitivity ca. 30 times larger than for the univariate reference method.