Análise discriminante do perfil de propriedades leiteiras de acordo com padrões de contagem de células somáticas e contagem padrão em placa

Abstract

The implementation of Good Agricultural Practices (GAP) holds extreme importance and refers to the adoption of adequate procedures throughout the whole production cycle, ensuring the milk is sustainably produced by healthy animals, being these some of the requirements of normative 77/2018. The study's goal was to draw the dairy farms' profile and to reduce the variables used to discriminate producers whose quality standards are in adequacy with IN 76 from those who are not, through the Fisher Discriminant Ratio (FDR) analysis. In order to do so, secondary data collected transversally by the company Cia do Leite has been used, divided into two data banks: 1. Milk quality standards - a. Somatic Cell Count (SCC) and b. Standard Plate Count (SPC) and 2. Interview results - a form applied to producers affiliated to the contracting dairy, in a single visit, between June of 2019 and February of 2020. The banks have been connected by producer's name. Descriptive analysis has been done of all raised variables to draw the interviewer’s and their properties' profile. SCC and SPC parameters have been categorized as "Acceptable" and "Unacceptable" under Normative Instruction 76/2018 by the Brazilian Agriculture, Livestock and Supply Ministry (MAPA). SCC numbers below 500k SC/mL, as well as SPC numbers below 300k UFC/mL, are "Acceptable", whereas numbers above these standards are "Unacceptable". FDR analysis has been done to distinguish two groups in two situations: 1. Of the entire data bank to discriminate "Acceptable" and "Unacceptable" categories; 2. Of a data subset, selecting properties with the 10% highest and lowest SCC and SPC parameters in order to determine which variables best discriminate the best and worst farms in the sample. Furthermore, the accuracy between the models has been tested through Decision Trees. It has been concluded the 5 variables which discriminated producers the most regarding accordance with SCC and SPC values are related to causal hygiene matters, presence of periodic technical assistance and technology implementation. The accuracy through discrimination does not change substantially when 5 or 50 variables are used and is slightly higher than 50% in all evaluated scenarios. This study shows a path to variable reduction for the discrimination of farms, regarding sanitary milk quality, which requires further study.