Study on properties and composition of ancient glass based on PCA and gradient lifting tree model

Abstract

In order to help archaeologists better study ancient glass, uncover the mystery of the early Silk Road, and also better protect these cultural heritage.In this paper, the properties and composition of ancient glass are studied.All types of glass samples were analyzed by PCA principal component analysis and weighted summation, and the type characteristic equation was obtained. After calculating the candidate points iteratively in MATLAB, the optimal critical eigenvalue was set to 5, and the type judgment model was established.According to get the type of the characteristic equation, the selection is related components as high potassium and the composition of class division, choose negative correlation components as lead, barium and the composition of class division, through the PCA principal component analysis of two types of glass data are reduced to 3 d in the class we USES Kmeans clustering algorithm for two types of glass, For the sensitivity analysis of subclass partitioning, we used SOBOL global sensitivity analysis method, and obtained that principal component 1 would have the greatest impact on the model output.