Support vector machine as tool for classifying coffee beverages

Abstract

Classifiers are tools widely used nowadays to process data and obtain prediction models that are trained through supervised learning techniques; there is a wide variety of sensors that acquire the data to be processed, such as the voltammetric electronic tongue, as a device employed to analyze food compounds. This paper presents a normal and decaffeinated coffee beverage classifier using a Support Vector Machine with a linear separation function, detailing the classification function and the model optimization method; to train the model, the data measured by 4 electrodes of a voltammetric tongue that is excited by a predetermined sequence of positive pulses is used. In addition, the results graphically show the measurements obtained, the support vectors and the evaluation data, the values of the classifier parameters are also presented. Finally, the conclusions establish an acceptable error in the classification of coffee drinks according to caffeine presence at the sample analyzed. © Springer Nature Switzerland AG 2020.