Identification of Glass Artifacts Based on Decision Tree and k-Means Method

K-means algorithm not only has good processing ability of images and data sets, but also has good stability and scalability, and has good clustering effect. In addition, the clustering effect will not be affected by the order of data sets, and there is no need for corresponding constraints on the range of data sets. During the weathering process, ancient glass undergoes a large exchange of internal elements with environmental elements resulting in changes in its composition ratio, which affects the correct judgment of its category. In order to analyze the classification law of high potassium glass and lead-barium glass, a decision tree is used to determine lead oxide (PbO) as the characteristic element, and when lead oxide (PbO)5.26, it is classified as high potassium glass, and vice versa as lead-barium glass; on the On the basis of the two categories, considering that weathering and unweathering have certain influence on the subclass division, a two-layer subclass division is adopted, divided into four major categories, and the standard deviation statistics are conducted for the chemical components of each category, and several chemical components with larger standard deviation are screened out as characteristic elements, and the subclass division is carried out by kmeans clustering method, and finally a decision tree is used to find the chemical components that can be used as the basis for the division. The goodness of the decision tree is evaluated and analyzed by ROC curve to illustrate the rationality of the classification, and sensitivity analysis is performed by the relationship between the number of clusters and the depth of the decision tree to illustrate the reliability of the results.