Analysis and Identification of Ancient Glass Composition Based on the RF K-means++ Method

Accurate classification in the analysis of ancient glass products is often hampered by sample diversity and incomplete chemical composition data. This study introduces a method that combines Random Forest and K-means++ clustering models to enhance classification accuracy and efficiency. Experimental results indicate that optimizing input variables with the Random Forest model allows K-means++ clustering to differentiate categories more effectively. This confirms the method’s effectiveness in managing complex and incomplete datasets and highlights the potential of machine learning in cultural relic identification and material science. The approach from this study can also be applied to other fields, such as biomaterial classification and environmental sample analysis.