The main objective of the project is to develop improved prediction methods of mastitis in dairy cows. This will be achieved by combining state-of-the-art mathematical / statistical methods with access to large amounts of online data.
The project will perform data mining for optimal use of the large online data bases
generated in automatic milking systems, and to develop algorithms to analyze
the multivariate data for optimized diagnosis of mastitis cases. Data mining
and imputation will be applied because all variables that can be used for
diagnosis are not necessarily measured on all individuals and at all times,
thus creating lots of missing data. Advanced statistical methods, such as
dynamic linear models, will be applied because mastitis is a latent variable
that cannot be measured directly and because the data is multivariate and
highly collinear both in time and space. We expect that the results of the
research will significantly improve the possibilities for early detection of
mastitis cases in dairy cows and thus increase the chances for preventive
actions and consequently reduce the costs and potential antimicrobial
resistance associated with mastitis.