Article ID Journal Published Year Pages File Type
1181644 Chemometrics and Intelligent Laboratory Systems 2008 7 Pages PDF
Abstract

The possibility of using two different artificial neural networks architectures (multi-layer feed-forward, MLF-NN, and counterpropagation, CP-NN) for the classification of 255 durum wheat samples from Sicily (Italy) was investigated and the performances of the optimal models were compared both among each others and to those resulting from the application of traditional chemometric pattern recognition techniques. When considering predictive ability over an independent test set, counterpropagation NN performed best, being able to correctly predict about 82% of the external validation samples (the corresponding predictive ability for MLF-NN, LDA and QDA was 72.0%, 50.9% and 52.7%, respectively.

Related Topics
Physical Sciences and Engineering Chemistry Analytical Chemistry
Authors
, , , , , ,