The inspection and analysis of heavy mineral oil plays an important role in dealing with traffic-accident cases. In order to obtain accurate classifications of heavy mineral oils, we collected infrared and Raman spectral data for 120 samples of five kinds of heavy mineral oils, including gasoline engine oil, diesel engine oil, grease, gear oil, and hydraulic oil. We established a classification and discrimination model for heavy mineral oil by using a support vector machine (SVM) combined with a spectral-fusion method. The results showed the accuracy of modeling classification using single-spectrum data to be rather low. When we modeled and analyzed the data obtained from primary spectral fusion, the classification and recognition rates for the five heavy mineral oils were slightly better, with an accuracy up to 75%. However, modeling that used data from intermediate spectral fusion combined with principal component analysis achieved complete differentiation among the five heavy mineral oils, with feature extraction from the 26-dimensional matrix being the best, with an accuracy up to 100%. In summary, spectral-fusion data combined with SVM modeling analysis can achieve complete separation among heavy mineral oils. The method improves the efficiency of inspection and identification, which fulfills the goal of rapid and accurate inspection for frontline law-enforcement personnel. It also provides theoretical support and a reference method for relevant cases.