杭州老虎洞窑古陶瓷成分的化学计量学研究

Study on Sherds Excavated in Laohudong Kiln by Chemometrics Method

用支持向量机算法研究了与杭州老虎洞古陶瓷有关的两个断源、断代问题。作为化学计量学的一种新型分类算法,支持向量机在小样本问题上表现出良好的泛化能力,与特征选择方法结合,可以有效处理样本少,特征多的问题。本研究综合利用支持向量机、特征选择算法和其它化学计量学算法研究了杭州凤凰山麓万松岭附近的古窑遗址和“传世哥窑”的断源、断代问题,证明老虎洞窑与郊坛下窑产品截然不同,万松岭附近地面收集瓷片样本是老虎洞窑宋代地层的瓷片滑落所致,而“传世哥窑”样品可能是老虎洞窑元代时的产品。实验表明:支持向量机算法与化学分析相结合可以成为研究古陶瓷断源和断代问题的一种新方法。

Chemometrics methods such as support vector machine （SVM） algorithm and feature selection methods were applied to the study on sherds excavated at Laohudong Kiln in Hangzhou. As a new type of chemometrics method, SVM can effectively handle problems with small number of samples and many redundant and/or irrelevant features especially when combined with feature selection methods. Ancient ceramic can be distinguished by analytical chemistry methods of components measured. Now too few samples and too many chemical components are measured that they will make the chemometrics methods be overffiting and the model to distinguish the sherds be not confident. So SVM is combined with the backward search feature selection algorithm and other chemometrics methods like Fisher discrimination algorithm to treat the problems of place decision and period decision of sherds excavated in Laohudong Kiln and the handed down Ge Ware. Modeling results show that （ 1 ） sherds excavated in Laohudong Kiln is different from those in Jiaotanxia Kiln, （2） sherds collected in Wansong Mountain is from Laohudong Kiln, （3） the handed down Ge Ware might be the products in Yuan Dynasty. From these works, It is clear that combined with chemical analysis, SVM can be a new kind of useful chemometrics method to treat the problems of place decision and period decision of ancient ceramic.