Mastitis detection in dairy cows by application of fuzzy logic

The aim of the present research was to develop a fuzzy logic model for classification and control of mastitis for cows milked in an automatic milking system. Recording of data was performed on the University of Kiel's experimental dairy farm “Karkendamm”. A data set of 403,537 milkings from 478 cows was used. Mastitis was determined according to three different definitions: udder treatments (1), udder treatment or somatic cell counts (SCC) over 100,000/ml (2) and udder treatment or SCC over 400,000/ml (3). Mastitis alerts were generated by a fuzzy logic model using electrical conductivity, milk production rate and milk flow rate as input data. To develop and verify the model, the data set was randomly divided into training data (284,669 milkings from 319 cows) and test data (135,414 milkings from 159 cows). The evaluation of the model was carried out according to sensitivity, specificity and error rate. If the block-sensitivity was set to be at least 80%, the specificities ranged between 93.9% and 75.8% and the error rate varied between 95.5% and 41.9% depending on mastitis definition. Additionally, the average number of true positive cows per day ranged from 0.1 to 7.2, and the average number of false negative positive cows per day ranged from 2.4 to 5.2 in an average herd size for the test data of 39.7 cows/day. The results of the test data verified those of the training data, indicating that the model could be generalized.Fuzzy logic is a useful tool to develop a detection model for mastitis. A noticeable decrease in the error rate can be made possible by means of more informative parameters.