This study shows the potential application of a potentiometric electronic tongue coupled with a lab-made DataLogger device
for the classification of dairy products according to the type of milk used in their production, i.e., natural, fermented and
UHT milk. The electronic tongue device merged a commercial pH electrode and 15 lipid/polymeric membranes, which were
obtained by a drop-by-drop technique. The potentiometric signal profiles gathered from the 16 sensors, during the analysis
of the 11 dairy products (with ten replicate samples), together with principal component analysis showed that dairy samples
could be naturally grouped according to the three types of milk evaluated. To further investigate and verify this capability,
a linear discriminant analysis together with a simulated annealing variable selection algorithm was also applied to the electrochemical
data, which were randomly split into two datasets, one used for model training and internal-validation using a
repeated K-fold cross-validation procedure (with 64% of the data); and the other for external validation purposes (containing
the remaining 36% of the data). The multivariate supervised strategy used allowed establishing a classification model, based
on the potentiometric information of four sensor lipid membranes, which enabled achieving a successful discrimination rate
of 100% for both internal- and external-validation processes. The demonstrated versatility of the built electronic tongue for
discriminating dairy products according to the type of milk used in their production combined with its simplicity, low-cost
and fast time analysis may envisage a possible future application in dairy industry.
