Research on Glass Classification Based on Systematic Clustering and Logistic Regression

Aiming at automatic glass classification and determining the appropriate number of clusters, a classification method based on systematic clustering and logistic regression is proposed in this paper. First, the elbow rule is employed to determine the appropriate number of clusters, and the results of glass subcategory classification are obtained by systematic clustering. Secondly, logistic regression model is built, which is fed into data to get the predicted probability value, and then the classification results of Binary Classification are obtained. Finally, the relationship between chemical composition and glass type is verified, and the sensitivity of the model is evaluated. Experimental results shows that the accuracy of our method rate reach up to 88% when the perturbation value is within 4%, which indicate that our model is high reliable for the classification of cultural relic glass and its chemical composition.