Composition Analysis and Identification Scheme of Ancient Glass Products Based on K-Means Algorithm

Abstract

Chinese ancient glass mainly includes high potassium glass and lead barium glass. In the process of weathering, its chemical composition is easy to change, affecting the judgment of its category. The data of this study are based on the chemical composition ratio of classified glass relics obtained by preprocessing. Firstly, a random forest model is constructed based on the Gini coefficient to obtain the contribution of lead oxide to the classification. The decision tree in the random forest is extracted, and the optimal classification rule is obtained by using the majority voting mechanism : when ω ( PbO ) is 10 %, it is lead barium glass. Secondly, the K-means clustering algorithm is divided into three sub-categories. The characteristic quantity with large standard deviation is used as the basis for classification, and the classification results are obtained : high potassium glass sub-categories include low calcium type, medium calcium aluminum type and high calcium aluminum type ; lead barium glass subclass includes low barium type, medium lead barium type, high lead barium type. For the identification of unknown types, the Euclidean distance between the unknown category and the sub-category centroid is calculated, and the minimum distance is taken to determine the sub-category it belongs to.