摘要
将多种单分类器模型融合,并用融合后的模型对不同品种干红葡萄酒进行判别分析。用 BRUKER MPA傅里叶变换型近红外光谱仪采集170个干红葡萄酒样品的近红外透射光谱,选取 PLS-DA,SVM, Fisher和 AdaBoost作为单分类器建模方法,分别建立葡萄酒品种判别模型,通过差异性度量值对单分类器进行筛选,得到差异性较大的四个单分类器作为基分类器,其中基分类器对测试集葡萄酒品种判别准确率最高为88.24%,最低为81.18%。然后通过加权投票机制对基分类器进行融合,融合后的模型对测试集葡萄酒品种判别准确率提高至92.94%,误判样品个数由单分类器最少的9个降为6个。实验结果表明多分类器融合所建立的模型优于传统近红外光谱定性分析一般采用单分类器模型结果,提高了葡萄酒品种判别的准确性,采用基于近红外光谱的多分类融合方法对葡萄酒种类判定具有可行性。
学科与关键词
关键词
葡萄酒多分类器融合差异性度量近红外
- Discrimination of Varieties of Apple Using Near Infrared Spectra Based on Principal Component Analysis and Artificial Neural Network ModelHE Yong; LI Xiao li; SHAO Yong ni光谱学与光谱分析, 2006, 26(5).
- Authenticating white grape must variety with classification models based on aroma sensors, FT-IR and UV spectrometryRoussel, S; Bellon-Maurel, W; Roger, JM; Grenier, PJournal of Food Engineering, 2003, 60(4): 407-419.被引 92
- Study on Fast Discrimination of Varieties of Acidophilous Milk Using Near Infrared SpectraHE Yong; FENG Shui juan; LI Xiao li; QIU Zheng jun2006.
