Based on the CART Decision Tree Model of Prediction and Classification of Ancient Glass-Related Properties

Abstract

Weathering occurs when glass interacts with the atmosphere. Ancient glass is susceptible to the influence of the burial environment. In the process of glass weathering, its composition ratio will be changed due to a large number of internal elements and environmental elements exchange. In this paper, Fisher's exact test and frequency analysis are used to test the glass pattern, type, color, and surface weathering respectively, and the conclusion is that there are significant differences between glass type and surface weathering, while there are no significant differences between pattern, color, and surface weathering. Then, the CART decision tree algorithm was used to classify the glass into weathered and unweathered, and a binary tree was established with 14 chemical components as characteristics. The average chemical composition of all the sampling points in the sibling nodes of the node where the weathering point is located was used as the predictive value of the chemical composition before the weathering point. Finally, the correlation analysis of the chemical constituents of high potassium glass and lead barium glass was carried out by hierarchical clustering method. In a high-potassium glass, silica is one category, and other chemical components are another category. In a lead-barium glass, silica and lead oxide are one category, and other chemical components are another category. Therefore, it can be considered that the chemical composition relationship of lead-barium glass is more complex.