Prediction and classification of chemical composition of ancient glass objects based on generalized Shapley functions

Abstract

Ancient glass products have suffered from the baptism of time and experienced changes in the burial environment and weathering, resulting in a change in the proportions of their chemical composition and interfering with their accurate identification by later generations. In this paper, the chemical composition of ancient glass products is predicted and identified. First, the multivariate statistical ANOVA test is applied to explore the relationship between whether the cultural relics samples are weathered or not and the glass type, decoration, and color to derive a law of chemical composition of the cultural relics and to analyze the correlation and difference among the four factors. Second, compared with the relevant data of the existing glass products, the missing values are processed by using the method of filling in the plurality. The weathering condition of the sampling points of the samples whose surfaces are not weathered is judged by the "distance discrimination method." Combined with the characteristics of the lead-barium glass and the high-potassium glass, the law of the chemical composition content on the surface of the samples, weathered or not, is explored. The modeling of the gray prediction method was applied again to predict the chemical composition content before weathering. Finally, the generalized Shapley function of fuzzy measurement was used to analyze the correlation between indicators and the chemical compositions and their differences. The scheme proposed in this paper can solve the difficult problem of category judgment in archeology, which is of great significance in promoting the smooth progress of archaeological work.

Keywords: ancient glass; correlations; generalized Shapley function; gray prediction; variability.