Classification of olive cultivars using artificial neural networks
Conference Paper
Overview
Overview
abstract
Olive fruit classification according to their cultivar is of major importance to guarantee varietal authenticity. Usually, non-supervised or supervised statistical tools (such as Principal Component Analysis or Linear Discriminant Analysis, respectively) are used for this purpose, based on several physico-chemical data, namely table olive fatty acids profiles, dietary fiber, sugar, organic acids and mineral nutrient contents. In this work, quantitative morphological parameters of fruit and endocarp were evaluated. Seventy samples, containing each one around 40 olives, of the six most representative olive cultivars of Portuguese northeast region (Cobrançosa, Cordovil, Madural, Negrinha de Freixo, Santulhana and Verdeal Transmontana) were selected. The samples were collected in different groves and during four crop years. The biometrical data was used together with a Multi layer Perceptron Artificial Neural Network allowing the implementation and validation of a classification model. Its performance was compared with that obtained using a linear discriminant analysis. The best results were obtained using artificial neural networks, especially for the external validation procedure implemented. The satisfactory results achieved, even when compared with previous published works, regarding olive cultivar's classification, show that the neural networks could be used by olive oil producers as a preventive and effective tool for avoiding adulterations o f Protected Designation of Origin or monovarietal olive oils with olives of non-allowed cultivars.
