Geographical recognition of Syrah wines by combining feature selection with Extreme Learning Machine. (May 2018)

• Record Type: Journal Article
• Authors: da Costa, Nattane Luíza; Llobodanin, Laura Andrea García; de Lima, Márcio Dias; Castro, Inar Alves; Barbosa, Rommel
• Is Part Of: Measurement. Volume 120(2018)
• Page Start: 92
• Page End: 99
• Publication Date: 2018-05
• Subjects: Extreme learning machine -- Artificial neural networks -- Support Vector machines -- Feature selection -- Wine classification; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2018.01.052
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: Extreme Learning Machine was used to classify Syrah wines. Classification accuracy of above 98% was obtained. ELM outperformed SVM and MLP. Abstract: Data mining techniques have been used for the classification of many types of products. In order to classify the Syrah wines from Argentina (Mendoza) and Chile (Central Valley), according to their origin, we perform two feature selection methods with the following classification algorithms: Support Vector Machines (SVM), and two types of artificial neural networks, Multilayer Perceptron (MLP) and Extreme Learning Machine (ELM), on 10-fold cross-validation. Each feature selection method has a different approach, creating also different sets of the most important features. The best model was the combination of variables peon-3-glu, malv-3-glu and pet-3-acetylglu, selected by Random Forest Importance, reaching 98.33% accuracy with ELM, outperforming SVM and MLP. The results obtained from the classifiers and feature subsets are able to confirm the importance of the anthocyanins to classify Syrah wines according to their geographic region. ELM was the best algorithm for classifying Syrah wines.