Glass Composition Analysis and Identification based on Improved Decision Tree Algorithm

Abstract

Glass is a valuable material evidence of our early trade exchanges, and our glass is mainly divided into high potassium glass and lead-barium glass. The specific types can be distinguished by examining the content of their internal compounds. This paper delineates the broad types of each unknown artefact and then predicts the composition before weathering. Combining the results of existing predicted subclasses as known data, 80% of the data is selected as the training set and 20% as the test set. Suitable node splitting evaluation criteria and feature division point selection criteria were determined, and a glass component analysis and identification model based on an improved decision tree algorithm was established to predict the major unknown artefact types separately and obtain their specific subclasses. The sensitivity analysis of the established identification model is carried out by adjusting the proportion of training data as well as the principal component weights.