Classification of ancient glass artifacts based on logistic regression and k-means clustering

Abstract

Chinese glass products have a long history. In this paper, for the weathering process and heritage components of ancient glass, we firstly established a logistic regression model to analyze the role of each class of chemical components on glass classification; then, in this paper, we identified the key elements with the greatest influence on glass classification: potassium oxide and lead oxide by gradually adding regression elements; subsequently, for each class of high potassium and lead barium glass, we Then, for each type of glass of high potassium and lead-barium, we selected some chemical components to establish a K-means clustering model, and after calculation, the high potassium glass was divided into four subclasses of GJ-1, GJ-2, GJ-3, and GJ-4, and the lead-barium glass was divided into four subclasses of QB-1, QB-2, QB-3, and QB-4, and the characteristics of each subclass were described. Finally, reasonableness and sensitivity tests were conducted to prove the reasonableness and robustness of the classification results.