A Statistical Method for Component Analysis and Identification of Differentiated Glass Products

Abstract

As the most widely used product in human daily life, glass was used as important material evidence for trade in ancient times. With the continuous development of my country's archaeological undertakings, the analysis and identification of the components of ancient glass products have become a research topic of practical significance. In order to predict the chemical composition content before weathering, we first analyze the relationship between surface weathering and the three variables of glass type, texture, and color from two aspects: correlation and difference. According to the discrete distribution state after quantification of qualitative data does not meet the normal distribution, Spearman correlation analysis and chi-square test model are used to draw the conclusion: relatively speaking, the degree of correlation between glass surface weathering and glass type is relatively high; glass type is closely related to the other three There are significant differences among the variables, decoration, and color. From the analysis of the influence of chemical composition, combined with relevant chemical data and verification, we can draw the following conclusions. In a certain range, the increase of CaO content is beneficial to increase the stability of the glass. From the point of view of numerical fitting solutions, we consider that the ridge regression model prediction model has higher sensitivity than other models such as Lasso regression due to the large difference in component values. Therefore, in this paper, we use the ridge regression model to fit the functional relationship between weathering and different categories of attributes and then use the prediction model obtained by ridge regression to solve the content before weathering.