Applications of Support Vector Machine in Tobacco Blend Grouping of Chinese Type Cigarette

Purpose Our aim was to solve the inconformity of cigarette formulated grouping with quality indexes.Method The support vector machine (SVM) was applied to learn the law that grouped with quality indexes. And then, the model was established and distinguished the divergent samples.Results The model could determine the divergent samples' belonging well, and the optimal c value was 2, g value was 0.5 by parameter optimization and 4-fold cross-validation, and the accuracy to prediction sets was 100%. Using the grouping method, distinct differences were found between the two modules, whether the sensory quality or aroma substance content. With the SVM, it shonwed more scientific than those methods relying on a single character.Conclusion The SVM method can reduce the subjectivity and one-sideness, and enhance the functional characteristics of the tobacco blend grouping. Therefore, the method provides a theoretical reference for the tobacco blend grouping.