Multi-agent working together based on the improved k-means

Aiming at the composition analysis and identification methods of ancient glass products, this paper proposes a clustering method based on Topsis and K-means. It uses the Topsis method to divide the subcategories of ancient glass based on K-Means unsupervised classification. Among the high- potassium glass types, the composite score index of silica (SiO2), potassium oxide (K2O), alumina (Al2O3), and calcium oxide (CaO) was higher. Among the lead-barium glass types, the composite score index of silica (SiO2), lead oxide (PbO), barium oxide (BaO), and phosphorus pentoxide (P2O5) was higher. The top four elements with the total score were selected, k-means clustering was used to subclassify ancient glass artifacts, and the model’s sensitivity was well-known by linear regression. It points out a new method of analyzing and identifying glass products, which is of great significance for investigating the exchange and dissemination of ancient glass objects and technologies.