As the basal group of Polypodiales, the specific taxonomy of Dicksoniaceae is still being debated. As a quantitative analysis method, numerical taxonomy has been applied to the taxonomic study of many plant families and genera in recent years due to its simplicity and high accuracy. However, the numerical analysis of the Dicksoniaceae fossils has not been reported at present. In the present study, the pinnule morphological data of 42 Mesozoic fossil species of the Dicksoniaceae were analyzed using cluster analysis, principal component analysis and correlation analysis. The results revealed that 42 taxonomic units could be divided into six representative groups, which are consistent with the traditional taxonomy. After screening, an identification key on 28 fossil species of four genera with a definite taxonomic position was established. According to the quantitative analysis, a Bayes discriminant model was established for the selected species. Lastly, the model was tested using the morphological data of the fossil pinnules in Dicksoniaceae from the Yaojie Formation, suggesting that the discriminant model is accurate to a certain extent. As a result, the numerical taxonomy can be applied to the classification of the Dicksoniaceae fossils.