Multivariate data analysis as a discriminating method of the origin of wines
| Autor: | Forina, M., Armanino, C., Castino, M., Ubigli, M. |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2015 |
| DOI: | 10.5073/vitis.1986.25.189-201 |
| Popis: | A data set of 178 wines from Piedmont (Barbera, Grignolino, Barolo) was evaluated by multivariate data analysis in order to both build the category models and single out anomalous samples. By feature selection (Fisher weights) only 8 variables, out of the 28 chemical and physico-chemical original variables of the data set, were selected on account of their high univariate discriminant ability. Classification methods (KNN, LDA, PCA) and modelling techniques (Bayesian analysis, SIMCA) were applied to the 8-dimension data set; classification ability was about 98 %. VITIS - Journal of Grapevine Research, Vol. 25 No. 3 (1986): VITIS |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|