Classification of cow milk using artificial neural network developed from the spectral data of single- and three-detector spectrophotometers

Abstract

Spectra data from two instruments (UV–Vis/NIR and FT-NIR) consisting of three and one detectors, respectively, were employed in order to discriminate the geographical origin of milk as a way to detect adulteration. Initially, principal component analysis (PCA) was used to see if clusters of milk from different origins are formed. Separation between samples of different origins were not observed with PCA, hence, feed-forward multi-layer perceptron artificial neural network (MLP-ANN) models were designed. ANN models were developed by changing the number of input variables and the best models were chosen based on high values of generalized R-square and entropy R-square, as well as small values of root mean square error (RMSE), mean absolute deviation (Mean Abs. Dev), and –loglikelihood while considering 100% classification rate. Based on the results, whether the spectra data was collected from a single or three detector instrument the same clustering was observed based on geographical origin.