Tracing the causes of macular cigarettes have been significant issues in the field of cigarette production. Through molecular fluorescence spectroscopy experiments, the feasibility of using fluorescence spectroscopy to recognize different kinds of macular cigarettes was verified. Based on these findings, a method combining fluorescence hyperspectral with machine learning algorithms was proposed to detect and recognize macular cigarettes. By using the spectral feature extraction methods Principal Component Analysis (PCA) and Successive Projection Algorithm (SPA), and combining several classification models, a comprehensive comparison shows that after processing the original fluorescence hyperspectral data and using the Random Forest (RF) model, the identification of different types of macular cigarettes can achieve the highest accuracy, with a rate of 95.36%. This indicates the feasibility of identifying different types of macular cigarettes using molecular fluorescence spectroscopy and exploring their causes and traceability. The established scheme holds significant application value for developing fast, non-destructive, and intelligent automatic identification equipment for the cause and traceability of macular cigarettes.