Abstract
In this paper, poly(acid chrome blue K) (PACBK)/AuNP/glassy carbon electrode (GCE), polysulfanilic acid (PABSA)/AuNP/GCE and polyglutamic acid (PGA)/CuNP/GCE were self-fabricated for the identification of rice wines of different brands. The physical and chemical characterization of the modified electrodes were obtained using scanning electron microscopy and cyclic voltammetry, respectively. The rice wine samples were detected by the modified electrodes based on multi-frequency large amplitude pulse voltammetry. Chronoamperometry was applied to record the response values, and the feature data correlating with wine brands were extracted from the original responses using the 'area method'. Principal component analysis, locality preserving projections and linear discriminant analysis were applied for the classification of different wines, and all three methods presented similarly good results. Extreme learning machine (ELM), the library for support vector machines (LIB-SVM) and the backpropagation neural network (BPNN) were applied for predicting wine brands, and BPNN worked best for prediction based on the testing dataset (R (2) = 0.9737 and MSE = 0.2673). The fabricated modified electrodes can therefore be applied to identify rice wines of different brands with pattern recognition methods, and the application also showed potential for the detection aspects of food quality analysis.