Mineração de dados para estimativas de mortalidade pré-abate de frangos de corte

Abstract

The performance of a simple information system to identify possible mortality conditions in pre-slaughter operation through classification and regression trees was studied. Data referring a commercial broiler slaughterhouse, located in State of São Paulo, were studied using a data mining system (CART). This decision tree is able to predict categorical variables through a historical dataset and then classify effects. The classification tree had 77 % of success and showed to be a practical and faster way to predictions for this dataset and it is able to show an easy interpretation to direct use by the sector area experts.