Determination of Glass Artifacts Based on K-means Clustering Algorithm and Support Vector Machine Modeling

The aim of this study is to classify and identify ancient glass artifacts through K-means clustering algorithm and support vector machine (SVM) model. The field of cultural heritage protection is expanding, and the study of weathering and chemical composition of ancient glass artifacts is becoming more and more important. However, traditional classification methods often rely on manual judgment and experience, and are difficult to handle complex chemical composition data, so this paper introduces a variety of machine learning algorithms to improve classification accuracy and generalization ability. Firstly, a list table is constructed and Pearson chi-square test is performed to analyze the relationship between the degree of glass weathering and the type. Then, to deeply explore the weathering characteristics of glass artifacts, statistical indicators (e.g., mean, standard deviation, skewness, and kurtosis) were applied to analyze the volatility of chemical composition. Then, in the establishment of the classification model, K-means clustering algorithm was applied to further subdivide the high-potassium and lead-barium glasses. Meanwhile, the average chemical composition trend of each type of glass is derived by polynomial fitting method and the chemical content before weathering is estimated. Finally, a kernelized soft-interval SVM model is introduced to build a glass type classifier characterized by chemical composition, which improves the classification accuracy. The stability of the decision tree model was verified through sensitivity analysis to ensure the reliability of the classification results. This research not only provides new algorithmic support for the scientific classification and identification of ancient glass artifacts, but also provides a strong technical guarantee for the application in the field of cultural heritage protection.