Identification models for glassware are important because the influence of the burial environment on the degree of weathering of old glass can lead to an inaccurate assessment of its category. This paper explores the relationship between glass surface weathering and decoration, color and type by Chi-square test and Fisher’s exact test. The experiment results show that the surface weathering degree of glass is not related to its pattern and color, but is significantly related to its type. This paper has introduced a new method to obtain important characteristic factors for distinguishing high potash glass and barium lead glass and for subclassification by Random Forest Model, so as to carry out R-type clustering to obtain classification results with 95% accuracy and high sensitivity. To identify the unknown category of cultural relics, the method of applying the BP neural network with the Bayesian regularization algorithm for training is proposed to improve the generalization ability of the model. In addition, based on the Spearman correlation coefficient, the differences in chemical composition correlation between different types of glass are compared in this paper. In conclusion, this paper explores a method for modeling and validation of glass component data and constructs an identification model that takes the accuracy of classification and prediction results and high robustness into account.