Fisher's linear discriminant analysis model for inscribing the effect of chemical composition on cultural heritage

Abstract

In this study, the relationship between weathering and glass type, decoration and colour was first analysed using Goodman and Kruskal's tau-y coefficients, and a chi-squared test was performed, which showed that weathering was most influenced by glass type, and was influenced by glass decoration but very little by glass colour. A Fisher linear discriminant analysis model was then established, and the eight chemical components with the greatest correlation to artefact type were calculated using SPSS software under the conditions of the two glass types respectively, and optimised using a grey correlation algorithm with specified thresholds for screening components, resulting in four final elements affecting high potassium glass as potassium oxide, magnesium oxide, iron oxide and copper oxide, and affecting lead and barium types as calcium oxide, aluminium oxide, lead oxide and barium oxide, aluminium oxide, lead oxide, barium oxide and strontium oxide.