Composition analysis and identification of ancient glass objects based on neural network models

Abstract

This paper presents a model based on a 3-layer feedforward neural network, which effectively preserves the characteristics of the chemical content of each category in ancient glass through 3 fully connected layers. The average prediction rate of the model was 96.43%, which was 2.43% higher than the traditional KNN classification model, 3.42% higher than the support vector machine (SVM) model and 8.43% higher than the random forest model, demonstrating the efficiency of the model.