To analyze the main components of paper ashes and distinguish paper types， the experiment mentioned in this study prepared 30 brands of paper into paper ashes， using an X-ray fluorescent spectrometer to measure its main components. Using measurement data trained by support vector machine （SVM） classifier， the paper type and brand source were determined. The experiments accurately determined the main component data of 90 sets of paper ashes， and randomly and proportionally generated training and test sets. Using the MATLAB experimental platform， the best parameters c and g of radial-base core functions were determined by interactive testing method， and a support vector machine classification model was established. The reasons for the model misjudgment were analyzed using Pearson correlation coefficients. This study shows that an SVM classification model can effectively achieve sample classification， can be used to test the type of paper ashes and brand source， is beneficial to solve the court-science-related problems， and can provide assistance for police to collect physical evidence at a crime scene.