期刊
FOOD CHEMISTRY
卷 153, 期 -, 页码 414-423出版社
ELSEVIER SCI LTD
DOI: 10.1016/j.foodchem.2013.12.081
关键词
Wine; Multi-element determination; ICP-MS; ICP-OES; Neural networks; Multivariate statistics; Geographical origin
资金
- Slovenian Research Agency [P1-0034-0104, P1-0017]
Inductively coupled plasma mass spectrometry and optical emission were used to determine the multi-element composition of 272 bottled Slovenian wines. To achieve geographical classification of the wines by their elemental composition, principal component analysis (PCA) and counter-propagation artificial neural networks (CPANN) have been used. From 49 elements measured, 19 were used to build the final classification models. CPANN was used for the final predictions because of its superior results. The best model gave 82% correct predictions for external set of the white wine samples. Taking into account the small size of whole Slovenian wine growing regions, we consider the classification results were very good. For the red wines, which were mostly represented from one region, even-sub region classification was possible with great precision. From the level maps of the CPANN model, some of the most important elements for classification were identified. (C) 2013 Elsevier Ltd. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据