Research on Glass Classification and Recognition based on Support Vector Machines

Abstract

Glass artefacts undergo weathering and a significant exchange of internal elements due to burial in the ground. In order to classify the glass types, Principal Component Analysis was used to transform the existing chemical indicators into a small number of principal components. Subsequently, Support Vector Machine approach was used to classify the glass types to be predicted based on the selected principal components. K-means clustering algorithm was then used to classify the two glass types into subclasses on the basis of the selected chemical components and to perform sensitivity analysis. Using this model, the glass types of the samples to be tested can be identified.