Composition analysis and identification of ancient glass based on K-Means clustering

Abstract

The analysis and research on the composition of ancient glass is widely used in the reduction and identification of the composition of archaeological relics. In this paper, the relevant data of CUMCM2022 problem C is used to establish a mathematical model, and multiple linear regression, binary logistic regression and K-means clustering algorithms are used to realize the systematic analysis of the chemical composition of cultural relics, and solve the problem of predicting the classification of cultural relics. In the study of ancient glass classification, multiple linear regression and binary logistic regression are used to transform abstract text data into intuitive multiple linear regression equation, and a relatively ideal classification law is obtained. According to the change rate of chemical composition before and after weathering of various kinds of glass, the appropriate chemical composition was selected as the classification index, and the elbow method and K-means clustering were used to obtain the distinctive subclassification results. Finally, the sensitivity of K-means clustering subclassification model is analyzed by numerical perturbation method, and the model shows high stability.