Subclinical Mastitis Detection by Machine Learning

Abstract

The study aimed at developing a robust Machine Learning algorithm that could accurately predict Subclinical Mastitis in dairy cows, which often goes undetected until it is too late. We utilized five Machine Learning models, including Decision Trees, K-nearest neighbours, Logistic Regression, Random Forests, and eXtreme Gradient Boosting, to find the best fit for the work. Data was collected for two years from two large Hungarian dairy farms using two databases: RISKA and ALPRO. Somatic Cell Count values and Electroconductivity of Milk variables were the key features used for Artificial Neural Network-based classification. By utilizing the features of the Caret Package in the R environment, we filtered out correlated explanatory variables.