Composition Analysis and Identification of Ancient Glass Objects Using Regression and Clustering Algorithms

Abstract

Among the artifacts excavated today, many are glass-products. Archaeologists have classified archaeological artifacts into two main types, high potassium glass and lead-barium glass, based on the chemical composition of the glass artifacts and other testing methods. For the protection of cultural relics, the chemical composition of the artifacts needs to be further analyzed and sub-categorized in order to achieve different protection measures for different categories. In this paper, firstly, based on the data, we analyze the classification pattern of high potassium glass and lead-barium glass, use logistic regression, test the significance of the results, and evaluate and analyze the regression results with the statistical distribution of the data. Then for different categories select suitable chemical indicators for subclass classification, suitable chemical indicators imply their relatively large amount of information, so use the entropy weight method to process chemical composition indicators to derive their information entropy weights, select chemical indicators according to the weights, and finally perform clustering classification according to the selected indicators. Finally, for the analysis of the rationality and sensitivity of the classification results, different classification algorithms may lead to different classification results, so the rationality and sensitivity can only be analyzed by comparing and analyzing the classification results based on different clustering algorithms for the selected chemical composition indicators.