Research on glass products classification based on machine learning

Abstract

In order to analyze the classification rules of ancient Chinese high-potassium glass and lead-barium glass and select the appropriate chemical composition for each category to divide them into subcategories, this paper models the type and chemical composition of glass by random forest algorithm, divides the data by 3:7 and uses 3-fold cross-validation, and uses Accuracy, Recall, Precision and F1-score to evaluate the model after training, and the results are optimal. The strongly correlated chemical components under each glass category were obtained by using the K-Means clustering algorithm, and the clustering results were visualized by using unsupervised learning, and the subclass division results were better when k was taken 2 after visualizing different cluster data. Finally, the results of subclassification are visualized and the rationality and sensitivity analysis of the method are carried out, and it is found that the method is accurate.