Component Analysis and Identification of Glass Products Based on Hierarchical Clustering and Naive Bayes

Abstract

In the process of weathering of ancient glass products, the interaction of internal elements and environmental elements will directly lead to the change of their composition proportion, thus affecting the judgment of the composition. In this paper, cultural relics were analyzed by chi-square test and spearman coefficient, and descriptive data analysis was used to process and predict the composition content of chemical data after weathering, and to preliminarily classify them with nuclear naive Bayesian method. The contour coefficient S was used to find the best K value, and the K-Means cluster analysis was used to select the chemical indicators of whether the different types of glass were weathered or not in 4 categories. Therefore, the best classification results were obtained respectively, and the relationship and difference between chemical components were obtained by Pearson correlation coefficient and principal component analysis. From the perspective of data analysis, this paper conducts statistical analysis for the data distribution law, uses K-Means and naive Bayes to divide the subclasses, and establishes the intelligent divisions of three different machine learning models after voting on the data. By combining the correlation analysis and the second reference number, the largest difference is the correlation of the chemical composition between the weathered high potassium according to the different European distance of the weathered high potassium, and the smallest difference is the correlation between the weathered lead barium and the unweathered lead barium.