Glass classification and identification based on logistic regression analysis and K-means++ clustering algorithm

Abstract

During the weathering process, a large number of internal elements are exchanged with environmental elements, resulting in a change in the proportion of glass composition, which affects the correct judgement of glass categories. By comparing the chemical composition of high potassium glass and lead-barium glass, this paper finds that the proportion of silicon dioxide and potassium oxide is significantly higher in the high potassium glass category, while the proportion of lead oxide and barium oxide is high in the lead-barium glass category, and the above-mentioned substances with large differences in content are defined as classified substances. A logistic regression analysis was carried out to derive a regression equation between the classified substances and the glass type, which was used as a basis for determining the glass type. Based on this, a K-means++ clustering algorithm was used to perform subclass classification, resulting in three subclasses for each type. Finally, the glass of unknown composition was identified according to the established model.