Aplicação de controle estatístico de processos e algoritmos de mineração de dados na gestão da produção de frangos de corte

Abstract

This work aims to apply techniques of statistical process control and data mining algorithms in the analysis, prediction and classification of productive variables that make up the characteristics of the birds produced by a cooperative located in the western region of Parana. The database provided by the company has a 2-year history of movement, containing the main production variables (Lineage, Average Weight, Productive Efficiency Index, Food Conversion, Feasibility, Lot Age and Daily Weight Gain), from the until the slaughter stage. In order to analyze whether the production processes of broilers, Coob and Coob Slow, met the cooperative's expectations, the average weight of 186 poultry lots was used. It has been found that the capacity indices are below one, which classifies the production processes of the aviaries as incapable. Three algorithms for data analysis (Multilayer Perceptron, Linear Regression and Naive Bayes) of WEKA software were used in the implementation of prediction and classification models. For this implementation, 6000 batches of birds of the Coob and Coob Slow strains were used, obtained between 01/07/2014 and 07/31/2016. The results show that the prediction models provide reliable estimates for the response variables: Average Weight and Productive Efficiency Index and the classification model shows a good performance in the classification of the bird strains.