In this paper, a BP neural network was established based on the deep learning framework to determine the category of ancient glass under the influence of weathering.Compared with random forest and support vector machine, the accuracy reached 100%, in which random forest and BP neural network predicted the same results. The convolutional neural network model was proposed to solve the classification problem, and the applicability of the convolutional algorithm in one-dimensional data was verified.Then, through the visualization of the chemical composition correlation of different categories of glass cultural relics, it is concluded that high-potassium glass has a strong correlation with silica, and lead-barium glass has a strong correlation with lead oxide.For the difference of chemical composition correlation between different types of glass cultural relics, the conclusion that silica in high-potassium glass is negatively correlated with other components was obtained by constructing knowledge map analysis.