Application of artificial neural networks in the geographical identification of coffee samples
2011
Borsato, Dionísio | Pina, Marcos Vinicios Roberto | Spacino, Kelly Roberta | Scholz, Maria Brígida dos Santos | Androcioli Filho, Armando
The recognition of the characteristics of coffee associated with a given agricultural system and aimed at adding value and attending the consumers’ demands stimulates the production of types of coffee properly described. The objective of this study was to explore and to explain the physicochemical characteristics and sensory attributes of the coffee grown in Parana State (Southern Brazil) based on an integrated approach of the terrior and the application of artificial neural network. Physicochemical variables of green coffee beans and roasted coffee beans were determined, as well as sensory attributes of the beverage. One hundred and seventy-two coffee samples were analyzed for moisture, proteins, chlorogenic acids, tannins, total acidity, total lipids, caffeine, total and reducing sugars and minerals. These properties were tabulated and presented to artificial neural network multilayer perceptron to be identified as the region and the city of planting. The artificial neural network classified correctly and tested 100% of the samples grown by region. For the database containing information by city, the automatic mode of the software Statistica 9.0 was used. The neural network showed 99% accuracy in training and 100% accuracy in the stage of testing and validation.
Показать больше [+] Меньше [-]Ключевые слова АГРОВОК
Библиографическая информация
Эту запись предоставил National Agricultural Library