Glass Cultural Relic Classification Based on Random Forest and Robust K-Means Algorithm

Abstract

The ancient glass relics have weathered to a certain extent under the influence of the environment, and there is a certain degree of chemical exchange between the glass relics and the external environment, which interferes with the classification of the glass relics. In order to better study the classification of glass relics. In this paper, the random forest algorithm is firstly used to determine the characteristic importance of various chemical components in high potassium glass and lead barium glass, and some chemical substances are sorted out, and then their classification rules are analyzed according to the selected chemical substances. Then, the elbow method was used to determine the number of two kinds of glass subclasses into 3 categories, and then the robust K-Means algorithm was used to divide the subclasses, and the characteristics of each subclass were analyzed and studied. Finally, the standardized mutual information NMI value was used to analyze the sensitivity, and the logic theory and classification results were used to evaluate the model.