Discrimination Of Milk Adulterated With Urea Using Voltammetric Electronic Tongue Coupled With Pca-Lssvm

This paper proposed a novel method for qualitative analysis of urea-tainted milk samples based on voltammetric electronic tongue coupled with multivariable analysis method. The electronic tongue adopted three kinds of metals, such as gold, platinum, and palladium, as working electrodes. 26 pure milk samples and 26 urea-tainted milk samples with different concentration values were prepared. The measurement principles involved the extraction of information from cyclic voltammograms recorded in unadulterated and adulterated milk samples. The response current for different working electrodes has been considered for data analysis. A mean-centering and normalization method was employed for the raw current data. The classification algorithm of this system is separated into two phases: (1) the feature reduction by principal component analysis (PCA); (2) the classification by least square support vector machine (LSSVM). When principal component factors are equal to 10, the optimal PCA-LSSVM model was obtained and discrimination rate of the LSSVM model are 100% and 94.12% in the calibration set and the prediction set, respectively. The results demonstrated that the voltammetric electronic tongue technology with PCA-LSSVM multivariable data analysis method can be successfully applied to the classification between unadulterated milk samples and adulterated milk samples.