Component analysis of ancient glass products based on hierarchical analysis clustering algorithm

Abstract

First, the attachment is pre-processed, and after the abnormal data are removed side by side, the bar chart is drawn to preliminarily analyze the relationship that lead-barium glass is easier to weathering than high potassium glass. Then the chi-square test is carried out to find that whether the weathering of the glass cultural relics surface is related to the glass type of the cultural relics, but there is no obvious relationship with the decoration and color of the cultural relics. Secondly, the statistical model of one-way ANOVA was established, and the difference analysis of various chemical components was conducted before and after the two types of glass weathering. The chemical components that passed the difference test had significant statistical rules. Finally, multiple linear regression equations are constructed to predict the chemical composition content before weathering. The 14 chemical components of glass were regarded as 14 indicators, and the principal component analysis method was used to calculate the principal component contribution rate and the cumulative contribution rate, and then the two principal components were determined. Then the principal component analysis was used to cluster the indicators, and the hierarchical clustering algorithm was used to generate lineage maps. According to the elbow principle, the number of subcategories of high potassium glass is 3, and the number of subcategories of lead-barium glass is 4, so as to divide the chemical composition of glass. The cluster group scatter plot is then plotted and the subclass results are highly reasonable. Then a stepwise regression model is established to analyze the classification sensitivity and obtain several indicators with high sensitivity, which can be used for more targeted protection and restoration of unearthed cultural relics.