Analysis of glass artifacts based on hierarchical clustering and regression models

Abstract

Glass is a valuable evidence of the early friendly trade between China and the Western countries. In this study, the rationality and sensitivity of the developed model were first evaluated by analyzing the definitive and quantitative data of each type of glass artifacts and establishing a model to classify the glass artifact categories into subcategories, as well as categorizing unknown types of glass. Considering the discrete and discontinuous nature of the data, the chi-square test was chosen to investigate the correlation between glass type, decoration and color and surface weathering. Secondly, descriptive statistics of high potassium and lead-barium glass artifacts before and after weathering were conducted to obtain statistical patterns about the chemical composition content, and then the chemical composition content of glass artifacts before weathering was predicted based on the mean rate of change of the chemical composition content of high potassium and lead-barium glass before and after weathering, respectively. Finally, this study further explored the relationship between glass artifacts and the content of each chemical composition using Spearman's correlation coefficient to derive the basis for the classification of glass types, based on the hierarchical clustering method to classify the subclasses of high potassium glass and lead-barium glass with and without weathering, respectively, and used the elbow rule to derive the most appropriate number of subclasses for the above four types of glass, which endowed the model with rationality.