A research on subclassification of ancient glass based on FLDA and systematic cluster analysis

Abstract

For the purpose of subclassification of ancient glass, this study first used FLDA (Fisher Linear Discriminant Analysis) to classify the glass, in this study, unknown glasses of known composition were reasonably classified by model analysis with a high accuracy of 92.5%, and the classification pattern of the two types of glass was analysed statistically in conjunction with the data, and the results showed that the high potassium type contained high levels of K and Na chemical components; the lead-barium type, on the other hand, contains a high content of Pb and Ba chemical components. Next, a systematic clustering model was used for cluster analysis, avoiding the disadvantages of the K-means clustering model which is sensitive to outliers, and further analysed by a combination of clustering dendrograms and the law of the elbow in the line graph of aggregation coefficient. The analysis was further Integrated analysed by the law of the Elbow in the clustering dendrograms and the line graphs of aggregation factors. The final subclasses were concluded as follows: high potassium glasses can be divided into two subclasses, K2O-CaO (~10wt%)-SiO2 and K2O-SiO2; lead-barium glasses can be divided into three subclasses, PbO-BaO-SiO2, PbO (~25wt%)-BaO-SiO2 and CaO-PbO (~40wt%)-BaO-SiO2. The study of the subclassification of ancient glass is very helpful and influential for industrial applications. Through the construction of the classification and clustering model in this study, the subclassification division of ancient glass is well carried out, which has an important theoretical basis to support the work of identification and detection, and is of great significance for the study of the origin of ancient glass.