粗糙集理论(RST)及其在古陶瓷分类上应用的初探

Introduction on rough set theory and its application to ancient ceramics classification

为了获得古陶瓷器的产地和制造年代的信息,需要应用科学分析方法对古陶瓷器的化学组成数据进行归纳、分类。本文介绍了粗糙集理论(RST)的一些基本概念,同时给出了用粗糙集方法建立古陶瓷分类模型的基本步骤和对古陶瓷器化学组成数据的处理方法,通过对瓷器产地、时代判别的应用举例,表明用该方法对古陶瓷分类研究的效果能达到预期目的,回代检验分析结果正确率为100％。

The chemical composition is usually acquired by various scientific analyses in Chinese ancient ceramics researches, then they are classified and catalogued by the calculation based on some proper mathematical model, finally the discrimination functions can be got. So the information of provenance and time for the unknown sample can be easily deduced. Rough set theory is a relatively new mathematical tool dealing with vagueness and uncertainty; it can be used to find hidden relations in data and to reveal latent rule of data. Rough set theory has been applied in many areas successfully. In this paper, the basic concepts of rough set theory were introduced, and a process of classifying ancient ceramics was set up primarily . A method to treat the data of the chemical composition of ancient ceramics also was gave, that is, constituting decision tables according to the data at first, then, acquiring a set of classification rules by reduction of condition attributes, reduction of values of condition attributes and reduction of decision rules. All data were analyzed by Energy dispersive X - ray spectrometer. Two examples to apply the method to acquire rules for the classification on the provenance and dating of ancient Chinese porcelain respectively were gave. One example was acquiring different rules for Ding kiln, Xing kiln, Gongxian kiln and Longquanwu Kiln, and another was for the blue - white glaze procelain of Yuan, Ming amd Qing dynasties. To verify the rules, all data back were inputed, the results showed that the rate of accuracy of the rules is 100% . It is illustrated that rough set theory could be used in acquiring knowledge about producing area and manufacturing time from composition database of ancient ceramics, and the method could acquire effectively a set of classification rules. In this paper, some other statistical analyses were also compared and their virtues and defects briefly were explained.