Herein, based on the fluorescence detection mechanism, acenaphthene, fluorene, and naphthalene were detected in polycyclic aromatic hydrocarbons by the parallel factor combined with the support vector machine (SVM) algorithm. The fluorescence spectral data were preprocessed and used as the training set, which was fed into the particle-swarm-optimized SVM algorithm to establish the classification model. The number of components was determined using the methods of core consistency analysis, residual square sum analysis, and iterative frequency analysis, and the optimal component number thus obtained was used to perform parallel factor decomposition. The obtained transmit load matrix was used as the test set and fed into the SVM classification model. The classification accuracy rate was 100%. The recovery rates of 100.45%±6.25%, 100.10%±6.39%, and 95.07%±7.46% were achieved for acenaphthene, fluorene, and naphthalene, respectively. The proposed algorithm avoids time complexity caused by human operation and errors caused by subjective factors. Thus, it can be applied for fluorescence detection of polycyclic aromatic hydrocarbons.