基于CatBoost的玻璃成分与风化关系实证探究  

An Empirical Study on the Relationship between Glass Composition and Weathering Based on CatBoost

作　　者：张星园 孟怡辰 罗建龙 李言 ZHANG Xingyuan;MENG Yichen;LUO Jianlong;LI Yan(North China University of Science and Technology,Tangshan 063210,China;Hebei Key Laboratory of Data Science and Application,Tangshan 063210,China)

机构地区：[1]华北理工大学,唐山063210 [2]河北省数据科学与应用重点实验室,唐山063210

出　　处：《玻璃》2023年第8期15-23,共9页Glass

摘　　要：高钾玻璃和铅钡玻璃的化学成分含量和物理属性不同,研究玻璃文物样品的类型对文物的保存和探索有着重要价值。为了精确分类玻璃文物样品类型,通过综合相关性分析研究两种玻璃的分类规律,发现玻璃类型和二氧化硅、氧化钾、氧化铅和氧化钡的相关系数可达到0.7,得知高钾玻璃中氧化硅和氧化钾含量较高,铅钡玻璃中氧化钡和氧化铅含量较高。结合样品的分类规律,以玻璃文物样品的化学成分以及表面风化情况作为特征参数建立CatBoost分类模型,并得到准确率为97%,召回率为95.9%,F1值为92.4%的模型效果,同时利用关联性和差异性分析进行模型验证,为考古工作者快速区分玻璃文物样品类型提供理论指导依据。High-potassium glass and lead-barium glass have different chemical compositions and physical properties.It is valuable to study the types of glass cultural relics for their preservation and exploration.For accurately classifying the types,the classification law of the two types of glass was studied through comprehensive correlation analysis.It was found that the correlation coefficient between the types of glass and silica,potassium oxide,lead oxide and barium oxide could reach 0.7 and the content of silica and potassium oxide were higher in the high potassium glass,and the content of barium oxide and lead oxide were higher in the lead barium glass.Combining the classification laws,the CatBoost classification model was established with the chemical compositions of glass artifact samples and surface weathering condition as characteristic parameters.The model effect with 97%accuracy,95.9%recall and 92.4%F1 value was obtained,and correlation and difference analysis were used for model validation.It provides theoretical guidance basis for archaeologists to quickly distinguish glass artifact sample types.

关 键 词：CatBoost算法 玻璃文物样品分类 综合相关性分析 差异性分析 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]