Subclass classification of ancient glassware based on K-Means and GMM

Abstract

This paper focuses on the composition analysis and type identification of ancient glass objects. This paper establishes a judgment model based on the surface characteristics of cultural relics samples, and analyzes the classification laws of high-potassium glass and lead-barium glass; then, the samples of the same category are classified into subclasses according to K-Means method, Gaussian mixture method (GMM) and Mini-Batch K-Means method, and according to each evaluation method, the K-Means method is chosen comprehensively to establish The K-Means method was used to establish a clustering model for glass samples, and the results of subclass classification under each category were obtained. Based on the composition and surface characteristics of the samples, descriptive statistics, scatter plots, S-W test and Pearson correlation coefficient were used to analyze the correlations of chemical compositions among samples of different categories; then, based on the subclass classification made in Problem 2, the differences of the correlations among samples of different categories of artifacts were analyzed using the methods of correlation analysis and difference analysis based on different ranges of data sets.