Component Analysis and Classification Study of Glassware Based on Random Forest and Decision Tree Classification

This paper addresses the problem of compositional analysis and type identification of ancient glass artifacts and establishes a decision tree classification model based on the basic characteristics and chemical composition content of glass artifacts. Firstly, the correlation between surface weathering and glass properties was analyzed by filling the plurality of missing information and Spearman's correlation coefficient method; secondly, the sensitivity of lead-barium and high-potassium compositional data was analyzed by mechanical learning classification using SPSSPRO software, and the data was accurately classified by the decision tree model; lastly, the weathering coded data was used by Randomforest method to prediction, and the sensitivity of the model was analyzed. In addition, the chemical compositions of different glass types were differentially analyzed by Spearman's correlation coefficient and Mann-Whitney test. This study provides a new methodology for the classification and identification of ancient glass artifacts and contributes a new perspective to the conservation and research of cultural heritage.