Glass Classification Accuracy Based on Spearman Coefficient and K-Means ++ Algorithm

The study of the physical and chemical properties of artifactual glass can contribute greatly to the attribution of artifacts and to the study of their history. In this paper, we study the interplay of physical and chemical properties of glass and classify them. In order to make the classification more accurate, the correlation coefficients of compounds were calculated using Spearman's coefficient, and the content of lead oxide was determined as the basis for the classification. After that, this paper used K-means++ to cluster the data and selected calcium oxide and alumina as the sub-classification basis for high-potassium glass, and lead oxide and barium oxide as the sub-classification basis for lead-barium glass. Finally, the contents of some artifacts calcium oxide, alumina, lead oxide and barium oxide were changed with a perturbation of 0.2%, and the clustering points did not move significantly, and the model was stable.