With the deepening of archaeological research, as a precious historical relic, the weathering of glass cultural relics has attracted more and more attention. In view of the effect of weathering on the chemical composition of glass cultural relics, two key questions are proposed in this paper: first, the chemical characteristics of the weathered glass before weathering are revealed by predicting the chemical composition of the weathered glass of lead-barium and high potassium type, that is, "composition reduction"; Secondly, a robust glass subclass classification model is constructed. The proposed model can not only accurately classify glass cultural relics when weathering causes the chemical composition to shift, but also ensure that the predicted deviation of the original composition of the weathered cultural relics does not significantly affect the classification results. By using the analytic hierarchy process (AHP) to select the key chemical components that can distinguish the glass sub-categories, the cultural relics are divided into four sub-categories: lead-barium - high sodium, lead-barium - low sodium, high potassium - low calcium and high magnesium, high potassium - high calcium and low magnesium. The corresponding classification model is established by support vector machine (SVM). The experimental results show that AHP model shows excellent classification accuracy and robustness for both unweathered data and weathered relics, which verifies its important application value in archaeological work.

Yifan Fu, Hang Guo, Liuhui Hong, An Gao, Wenxuan Zhou. Research on classification of glass types of ancient cultural relics based on support vector machine model. Academic Journal of Materials & Chemistry (2024) Vol. 5, Issue 3: 104-109. https://doi.org/10.25236/AJMC.2024.050316.

[1] Gao, G. & Wang, Q. Based on support vector machine (SVM) classification of ancient glass products [J]. Journal of jiangxi normal university, 2024, (3): 19-22. 

[2] Han Li, WANG Junyin. Composition analysis and type identification of ancient glass relics [J]. Journal of Yuxi Normal University,2024, 40(03):78-86.

[3] Shi Wei. Research on the composition of ancient Glass Products based on statistical analysis method [J]. Shandong business vocational college journal, 2024, 24 (02): 110-114. 

[4] Wang B G, Jiang J J. Identification of ancient glass species based on composition data analysis and fuzzy pattern recognition [J]. Science & Technology Innovation and Application, 2024,14(07): 41-46. 

[5] Qian H, Yue S, Xu L, et al. Ancient glass system discrimination method based on multivariate linear regression [J]. Journal of gansu science and technology, 2024, 40 (02): 56-61. 

[6] Shao Guangming, Xia Xianqi, Yin Hejie. Based on CART and cluster analysis classified prediction model of ancient glass research [J]. Journal of tonghua normal university, 2024, (02) : 31-35 

[7] Xie, H., Zhang, J., Guo, L. et al. Research on Mathematical Modeling of Properties and Chemical Composition of ancient Glass Products [J]. Journal of Dezhou University, 2023, 39(06):6-14.

[8] Chen, X., Ye, et al. Classification and identification of ancient glass cultural relics based on K-Means clustering and SVM algorithm [J]. Journal of Natural Science of Harbin Normal University, 2023, 39(04):70-79.

[9] Jiang, S., Chu, Z. l., Li, J. et al. Ancient relics glass chemical element correlation analysis [J]. Journal of chemical engineering and equipment, 2023, (7): 23-25. 

[10] Shi, B., Zhao, X., Wang, H. et al. Based on multilayer perceptron network of ancient glass products category prediction [J]. Journal of lanzhou liberal arts college journal (natural science edition), 2023, 5 (3): 57-62.