Abstract: To enhance tobacco origin and grade identification accuracy, a tobacco aroma recognition method combined with gas chromatography-ion mobility spectrometry（GC-IMS）technology and machine learning was proposed, to accurately distinguish tobacco origin and quality grade．Volatile organic compounds（VOCs）in tobacco leaf samples were qualitatively analyzed using GC-IMS, identifying 108 compounds, with characteristic peaks annotated for each compound were marked in GC-IMS spectrum．Differential analysis of compound intensities was further conducted to compare variations in compound intensity across tobacco from different provinces, prefecture-level cities, and quality grades．Results showed that significant variations in compound intensities across different origins and quality grades, providing valuable data for subsequent aroma modeling and quality assessment．A support vector machine（SVM）classifier was used to classify tobacco samples by origin and quality grade, achieving classification accuracies of 97.10% for provinces, 91.30% for prefecture-level cities, and 95.65% for quality grades．Ten random splits were performed to validate model's stability．Research demonstrated that combination of GC-IMS and SVM classification for volatile prganic compund identification effectively supported tobacco aroma analysis．An efficient tool for quality assessment and brand traceability was provided, while opening new research directions for GC-IMS application in food aroma evaluation．