Study of relic classification based on neural network and K mean classification

Abstract

Over time, its own chemical elements will change in response to environmental changes, commonly known as weathering, which will lead to changes in the chemical composition ratios and affect the judgment of the class. This paper focuses on the chemical composition of different types of glass. By comparing the chemical indicators of each type, an identification model of the cultural relic category is established to predict the chemical content of the weathering process. To solve this, the neural network, chi-square test, and Spearman correlation analysis are used and the calculation results are given. According to the definition of sensitivity, after analysis, the accuracy was taken as the evaluation index, and the type of chemical composition was taken as the factor. K-value cluster analysis was performed for the type of chemical composition of the two glass categories in turn, and the accuracy was compared, and the accuracy of high potassium glass was close to 83%, and the accuracy of lead-barium glass was close to 69%, concluding that the type of chemical composition, which has little effect on the accuracy, has little sensitivity. In different categories, the variation of independent variables has a large impact on the accuracy and a large sensitivity. Finally, the advantages and disadvantages of the model are analyzed, and the model is extended to a larger domain.