The weathering process of ancient glass causes changes in the ratio of its chemical composition due to the influence of the external environment, which affects the correct determination of its category. In order to classify the composition of glass products and thus analyze the variability between the chemical compositions obtained. This paper establishes a glass classification model based on K-Means++ , and uses the unsupervised learning K-Means algorithm to find the seeds of the center of mass of K-means clustering by heuristic method, and determines the optimal number of classifications by C-H value, and analyzes the clustering of different types of glass before and after weathering and the changes of the categories to which each chemical component belongs before and after weathering according to the clustering results. The CHI index of high potassium glass increased significantly after weathering, and that of lead-barium glass decreased relatively after weathering; for the DBI index, the DBI index of high potassium glass increased and that of lead-barium glass decreased after weathering; for the contour coefficients, the analysis showed that the chemical composition correlations were stronger after weathering for lead-barium glass and weaker after weathering for high potassium glass.

Zhengyang Wang, Wenbo Shen. Analysis and identification of the compositional correlations of glasses and the variability of different classes of compositions. Academic Journal of Materials & Chemistry (2023) Vol. 4, Issue 5: 25-30. https://doi.org/10.25236/AJMC.2023.040504.

[1] Mo Dingyuan. Research on the relationship between GDP and industrial structure in Baise City based on gray correlation analysis [J]. Mall Modernization, 2022(12): 124-126.

[2] Zhu Yingpei. Research on the composition system and production process of glass beads excavated from Yanghai Cemetery, Shanshan County, Xinjiang [D]. Northwestern University, 2018.

[3] Huang ZY, Zhou Lingke. Fault classification based on K-means Bayes and AdaBoost-SVM [J]. Computer System Applications, 2022, 31(07): 239-246.

[4] Kang Wei, Cao Wengeng, Xu Lixia, Nan Tian, Gao Yuanyuan, Nie Ziyi. A representative groundwater level calculation method based on k-means clustering and Tyson declustering [J/OL]. South-North Water Diversion and Water Science and Technology (in English and Chinese): 1-12 [2022-09-18]. http://kns.cnki.net/kcms/detail/13.1430.TV.20220726.1029.002.html

[5] Pu Huizhong. K-means cluster analysis algorithm in artificial intelligence + personalized learning system [J]. Intelligent Computers and Applications, 2022, 12(08): 152-156.

[6] Lin E Y. Starbucks as the Third Place: Glimpses into Taiwan's Consumer Culture and Lifestyles[J]. Journal of International Consumer Marketing, 2012, 24(1-2):119-128.

[7] Hu Yongkai. Big data visualization implementation under cloud platform [D]. University of Electronic Science and Technology, 2022.