CLASIFICADOR DE DATOS BIOLÓGICOS BASADO EN OPTIMIZACIÓN DE UN CONJUNTO DE REDES NEURONALES ARTIFICIALES VÍA META-LEARNING

Abstract

This work describes the development and optimization of the classier based on a Multi-layered Feed Forward Articial Neural Network (FFANN), used for the sorting of beef sample by age and detecting if the meat is older than thirty months (OTM). This classication is mandatory for the regulation of food security's matters in the majority of countries. The OTM classication is used with the purpose of oering security to the food chain, taking in to account that the bovine meat is under thrity months (UTM), this is a key indicator of the mad cow disease (bovine spongiform encephalopathy) absence. This classier is implemented as an FFANN trained through backpropagation, whose learning parameters (learning rate and momentum) and model (neurons in the hidden layer) have been optimized through Meta-Learning. Performing ensemble classication using the implementation of algorithms of ensemble methods in the framework. The performance of the classier is tested against others standar classiers with satisfactory results (SVM, SIMCA, PLS-DA and LDA).