In the processes of tobacco blend design and product maintenance of cigarettes, the evaluation of tobacco style and quality is very important. At present, the evaluation method mainly relies on artificial sensory analysis, which is subjective and relatively unstable. In order to realize the digital evaluation of tobacco style and quality, the pyrolysis and combustion characteristics of tobacco are taken as an efficient index, which is closely related to cigarette burning behavior and can be expressed by thermal analysis spectra. In this study, the thermal analysis spectra of 88 single-grade tobacco leaves were obtained by thermal analysis technology, and Support Vector Machine (SVM) algorithm was employed to classify 74 single-grade tobacco leaves with different qualities and styles into eight categories. The accuracy of the SVM algorithm for the training set was up to 98.6 %. Afterwards, the SVM algorithm was applied to predict the quality and styles of the other 14 single-grade tobacco leaves, and the accuracy was 92.9 % for the testing set. The results showed that this method could not only greatly improve the accuracy of digital evaluation of tobacco quality and style but could also verify the effectiveness and practicality of thermal analysis spectra for the digital evaluation of tobacco.