Improvement of accuracy in beer classification using transient features for electronic nose technology

Abstract

In this paper, an electronic nose technology was applied in beer classification. This paper proved the effectiveness of the use of odor for the classification of beers with different brands, which can imply a way for the characterization of beer using its aroma in the brewing industry. The method proposed in this paper was proved to be able to improve the classification accuracy very effectively. With transient features inclusive and dimensionality reduction by the principal component analysis, the accuracy of the classification was explicitly higher than a simple and traditional way of the use of only steady-state feature. As a result, the odor patterns of beer samples were clearly well-separated when including transient features. Furthermore, a simple-but-powerful learning vector quantization algorithm was used to derive the rate of classification accuracy. Consequently, it showed 100% of classification accuracy when including transient features, compared with 61% accuracy in the traditional method.