The current investigation analysed the applicability of support vector machines (SVMs), a sub-discipline in the field of artificial intelligence, for the early detection of mastitis. Data used were recorded on the Karkendamm dairy research farm (Kiel, Germany) between January 2010 and December 2011. Data from 215 cows in their first 200 days in milk (DIM) were analysed. Mastitis was specified according to veterinary treatments and defined as disease blocks. The two different definitions used varied solely in the sequence length of the blocks. Only the days before the treatment were included in the blocks. The following parameters were used for the recognition of mastitis: milk electrical conductivity (MEC), milk yield (MY), stage of lactation, month, mastitis history during lactation, deviation from the 5-day moving average of MEC as well as MY, and the 5-day moving standard deviations of the same traits. To develop and verify the model of the SVMs, the mastitis dataset was divided into training and test datasets. Support vector machines are tools for statistical pattern recognition, focusing on algorithms capable of learning and adapting the structure of the input parameters based on the training dataset. The results show that the block sensitivity of mastitis detection considering both mastitis definitions was 84·6%, while specificity was 71·6 and 78·3%, respectively. Showing feasible features for pattern recognition of biological data, SVMs can principally be applied for disease detection. However, without further performance improvement or different study settings (e.g. other indicator variables) SVMs cannot be easily implemented into practical usage.