Machine Learning-based Composition Analysis of Ancient Glass Objects

Ying Li, Jierong Tang, Junreng Rao, Yuhan Wang, Le Li, Zhen Tan, Weidong Xiao

Abstract: Ancient glass is one of the objects studied in archaeology, and the study of ancient Chinese glass has attracted much attention from scholars at home and abroad in recent years. Ancient glass is similar in appearance to exotic glass objects, but the chemical composition is different. Moreover, ancient glass is highly susceptible to weathered by the burial environment. During the weathered process, a large number of internal elements are exchanged with environmental elements, resulting in changes in their composition ratios, which affects scientists' correct judgement of their categories. This paper implements a variety of machine learning methods to analyze the chemical composition of ancient glass, predict the chemical composition content of ancient glass before weathered, correctly determine the type of ancient glass based on its chemical composition, and sub-classify ancient glass. The final results of this paper provide a scientific basis for archaeological research work and lay the foundation for subsequent studies on the origins, systems, manufacturing dates and preparation processes of ancient glass.