Composition analysis and identification of glass products based on classification and K-means + +
基于分类研判与 K-Means++的玻璃制品的成分分析与鉴别

摘要

古代玻璃在风化过程中，由于内部元素与环境元素进行大量交换，导致其成分比例发生变化，从而影响对其类别的正确判断，本文通过对已知数据进行统计性分析进行相关的分类研判并采用无监督学习的K-means++算法对玻璃制品成分进行分析与鉴别。首先进行数据预处理，根据题目要求，删除编号为15和17的两条错误数据， 将颜色为空值的部分根据风化程度规律设定为黑色处理，将其他空值进行填“0”处理，进行接下来的计算首先针对四个定类变量使用卡方检验进行分析，根据显著性P值是否小于0.05判断出玻璃类型和颜色与表面风化存在显著性差异，与纹饰不存在显著性差异，在此基础上进行效应量化分析，其中包含phi、Crammer's V、列联系数与 lambda。

Abstract

In the weathering process of ancient glass, due to a large amount of exchange between internal elements and environmental elements, the composition proportion of ancient glass changed, which affected the correct judgment of its category. This paper conducts relevant classification research and judgment through statistical analysis of known data, and uses K-means++algorithm without supervision to analyze and identify the composition of glass products. First, conduct data preprocessing, delete the two wrong data numbered 15 and 17 according to the requirements of the title, set the part with blank color as black according to the weathering degree law, fill in "0" for other blank values, and conduct the next calculation. First, use chi square test to analyze the four categorical variables, According to whether the significance P value is less than 0.05, it can be judged that there is a significant difference between glass type and color and surface weathering, and there is no significant difference between glass type and color and ornamentation. On this basis, quantitative analysis of effects is conducted, including phi, Crammer's V, contingency coefficient and lambda