古代玻璃制品的成分分析与鉴别
Composition Analysis and Identification of Ancient Glass Products
DOI: 10.12677/MOS.2023.121007, PDF,   
作者: 项 辉, 金 乐:上海理工大学机械工程学院,上海;王 可, 李霁昊:上海理工大学光电信息与计算机工程学院,上海;李 响:上海理工大学出版印刷与艺术设计学院,上海
关键词: Fisher判别分析皮尔逊相关系数RUSBoost算法决策树模型ROC曲线网格搜索法Fisher Discriminant Analysis Pearson Correlation Coefficient RUSBoost Algorithm Decision Tree Model ROC Curve Grid Search Method
摘要: 在经过初步分析后,使用SPSS中的Fisher判别分析对附件1的样本数据整体进行两次Fisher判别分析,得出化学成分中对于分类有明显影响的主要变量和高钾玻璃、铅钡玻璃的分类规律。分析得知风化文物的未风化采样点与未风化文物的采样点性质相似,故将其视为未分化组样本再使用excel进行统计性分析。首先使用SPSS分别画出高钾玻璃化学成分和铅钡玻璃化学成分的散点图,若判断具有线性关系,再使用皮尔逊相关系数进行不同类型玻璃文物化学成分相关性分析。考虑到风化作用的影响,所以将高钾玻璃和铅钡玻璃分为风化高钾、未风化高钾、风化铅钡、未风化铅钡4类进行分析。在进行未知文物的类型鉴别分析时,使用RUSBoost算法和决策树模型分别进行鉴别,将两种方法得到的结果使用ROC曲线进行比较得出类型鉴别结果更好的模型,并使用网格搜索法进行模型的敏感性分析。本文对于探究高钾、铅钡玻璃的分类规律以及玻璃文物的分类有一定的应用价值。
Abstract: After the preliminary analysis, the Fisher discriminant analysis in SPSS was used to conduct two Fisher discriminant analyses on the whole sample data of Attachment 1, and the main variables in the chemical composition that had a significant impact on the classification and the classification rules of high-potassium glass and lead-barium glass were obtained. After analysis, it is found that the unweathered sampling points of weathered cultural relics have similar properties to those of unweathered cultural relics, so they are regarded as undifferentiated group samples and then used excel for statistical analysis. Firstly, SPSS was used to draw scatter plots of chemical compositions of high-potassium glass and lead-barium glass respectively. If it is judged that there is a linear rela-tionship, then Pearson correlation coefficient was used to analyze the chemical composition correla-tion of different types of glass relics. Considering the influence of weathering, high-potassium glass and lead-barium glass are divided into four categories: weathered high-potassium, unweathered high-potassium, weathered lead-barium and unweathered lead-barium for analysis. RUSBoost algo-rithm and decision tree model were respectively used for the type identification analysis of un-known cultural relics. ROC curve was used to compare the results of the two methods to obtain a model with better type identification results, and grid search method was used to analyze the sen-sitivity of the model. This paper has certain application value to explore the classification rules of high potassium, lead barium glass and the classification of glass relics.
文章引用:项辉, 王可, 金乐, 李霁昊, 李响. 古代玻璃制品的成分分析与鉴别[J]. 建模与仿真, 2023, 12(1): 67-80. https://doi.org/10.12677/MOS.2023.121007

参考文献

[1] 黄利文. 基于变量择优的Fisher逐步判别分析方法[J]. 系统科学与数学, 2021, 41(8): 2338-2348.
[2] 张景香, 刘亚丹, 杨钊. 基于SPSS分析煤灰白度的影响因素[J]. 煤炭技术, 2022, 41(2): 230-231. [Google Scholar] [CrossRef
[3] 钟华星. 基于RUSBoost算法的违约风险预测模型构建与应用[J]. 财会月刊, 2020(10): 74-80. [Google Scholar] [CrossRef
[4] 樊良优, 姚小强, 王刚, 刘伟, 夏智权. 基于决策树的指挥控制决策方法[J/OL]. 电光与控制: 1-7.
http://kns.cnki.net/kcms/detail/41.1227.TN.20220913.1940.033.html, 2022-11-04.
[5] 宋花玲. ROC曲线的评价研究及应用[D]: [硕士学位论文]. 上海: 第二军医大学, 2006.
[6] 马小凤. 机器学习在学生通过率预测中的应用研究[D]: [硕士学位论文]. 重庆: 西南大学, 2019.