Research on glass type classification based on Logit model and K-Means clustering algorithm

Abstract

Glass artifacts have been important in human history and are often studied by archaeologists and art historians to understand the development of society, technological progress, and cultural exchange. However, the systematic classification of glass artifacts is a major challenge to their study, because glass artifacts excavated from archaeological excavations are often highly weathered, making it difficult to classify them, so a scientific and reliable method to analyze and systematically classify glass artifacts based on their detected chemical composition is of great importance to the study of human history and culture. In this paper, two methods, logit model classification and cluster analysis, were used to determine the key to distinguishing high-potassium glass from lead-barium glass by the content of PbO, SrO, SnO2, and CaO components. Next, a comparison of the three cluster analyses was used to determine the use of the k-means algorithm to further subdivide the high potassium class and the lead-barium glass artifacts into two subclasses each: high Al-Fe and high Cu-Zn; and high PbP and high Na-Zn. Finally, the sensitivity analysis of the model and the robustness of the model and the reasonableness of the results were analyzed using Pearson correlation coefficients.