A Study on the Classification of the Production Country of Korean and Chinese Traditional White Porcelain Using Machine Learning Models: Based on the Chemical Analysis Data of White Porcelain

Abstract

The objective of this study is to investigate the applicability of machine learning techniques in discriminating the production origins of pottery. 732 data sets analyzing the chemical composition of traditional white porcelain from Korea and China were collected, and models for determining the production country were developed by applying various machine learning algorithms. Upon applying these models to 146 test samples, the statistical analysis, principal component analysis-line discriminant analysis yielded a prediction accuracy of 87.7%, while machine learning techniques such as decision tree, K-nearest neighbor, and support vector machine models demonstrated relatively high prediction accuracies of 96.6%, 98.6%, and 99.3%, respectively. Additionally, feature importance analysis confirmed that rubidium is consistently the most critical variable for determining the origin in machine learning techniques that exhibit superior performance in classifying complex data structures. These findings underscore the potential of machine learning techniques in effectively discerning the production origins of white porcelain from Korea and China.