Machine Learning Can Suggest Management Decisions on Unperceived Mastitic Animals According to 1 Structural and Productive Profiles of Dairy Farms 2 3 Félix Goyache

14 15 In this paper the implementation of a tool was considered, based on Machine Learning (ML) techniques to improve 16 the detection of subclinical mastitis in dairy herds. The analysed population included 71 dairy farms in Asturias 17 (Northern Spain) and a total of 2,407 lactating cows. Four sources of information were available: a) a questionnaire 18 survey describing facilities, milking routines and management practices of the farms studied; b) dairy recording 19 information; c) classification of the cows suspected of being subclinical mastitic according to farmers' expertise; and 20 d) positive or negative scores with respect to the California Mastitis Test (CMT). The linear score (actual or 21 previous), lactation number, herd size, lactating cows per milker, urea concentration, number of clusters and 22 actual SCC are the most informative attributes for mimicking both farmers' expertise or CMT performance to 23 detect mastitis in dairy herds. However, to improve the detection of cows suspected of being subclinical mastitic, 24 the system uses other information related to management and milking routines, whereas it takes basically 25 productive information from the cow and its herd from dairy recording to classify a record as CMT-positive. 26 Decision rules inferred to predict CMT performance can provide useful, additional information to farmers to 27 improve the management of dairy herds included in milk recording programs where continual monitoring is 28 carried out. 29 30 1 dairy cow/subclinical mastitis/artificial intelligence/milk recording/somatic cell count 1 2