Analysis and Classification of Glassware based on Decision Trees and K-Means Clustering Model

In ancient times, different kinds of glass products appeared due to different combustion promoters added to the glass during the melting process. Glass is easy to be weathered during burial, resulting in changes in chemical composition. Therefore, it is difficult to analyze its composition and determine its type. Firstly, Pearson correlation coefficients were used to analyze and the correlation coefficients of surface weathering with type, color and decoration were obtained as:0.34, -0.22 and 0.05, respectively. Then the patterns of chemical composition contents were counted for different types of glass in different states, and it was obtained that barium glass showed a decreasing and increasing trend after weathering, respectively. Then a decision tree was used to predict the chemical composition content of the glass before weathering. Then, the binary statistical analysis was used to determine that when the total content of Na2 O, K2 O and CaO was greater than the total content of PbO and BaO, the glass type was Highpotassium glass, and the opposite was lead-barium. Then K-Means clustering analysis was used to obtain the subclassification of both High-potassium glass and lead-barium glass as three types. Then, using BP neural network, with all chemical components as input layer and glass type as output layer, we obtained Al, A6, A7 as High-potassium glass and A2, A3, A4, A5 as lead-barium. Finally, using Pearson correlation coefficients, the comparison found that the individual chemicals in High-potassium glass were more correlated.