基于PCA-BP神经网络的古代玻璃分类模型

Ancient glass classification model based on PCA-BP neural network

为了有效地鉴别古代玻璃并分析其主要成分,提出了一种预测古文物玻璃制品类型的方法。根据现有数据测量得到13种主要化学成分含量,基于主成分分析(PCA),将所得主成分作为反向传播算法(BP)神经网络的输入,构造一种基于PCA-BP神经网络的古代玻璃分类模型。实验中,选择80%的数据作为训练集与测试集,选择20%的数据作为验证集。结果表明:对玻璃文物样品提取的主成分有显著贡献的化学成分为SiO 2、K 2O、PbO和BaO;改进后的模型与传统神经网络模型相比,对样品预测平均相对误差率小于4%,迭代时间缩短,对未知的古玻璃文物样品的预测估计更精确;提出的玻璃分类模型在不同地区的不同数据集上有可靠的精确度,并相较于Logistics模型有较好的预测效果。

In order to effectively identify ancient glass and analyze its main components,a method was proposed to predict the type of ancient heritage glass products.Based on the existing data,13 main chemical components were measured,and the principal components were extracted based on the principal component analysis method,and the obtained principal components were used as the input of BP neural network,so as to construct a model based on principal component analysis and BP neural network for predicting the type of ancient glass objects.In the experiments,80%of the data were selected as the training set and the test set,and 20%of the data were selected as the validation set.The experimental results show that the chemical components that contribute significantly to the principal components extracted from the glass artifact samples are SiO 2,K 2O,PbO and BaO.The improved model,compared with the traditional neural network model,has an average relative error ratio of less than 4%for sample prediction,shorter iteration time,and more accurate prediction estimation for unknown ancient glass artifact samples,and can be used for the study of ancient glass artifacts.The proposed glass classification model can have reliable accuracy on different datasets from different regions and has better prediction results compared to the Logistic model.