Mastitis is one of the main diseases in dairy cattle, causing great economic losses to producers due to the reduction in milk production and changes in the quality of the final product. In this aspect, the identification of sick cows is essential for reducing the number of affected mammary quarters and reducing the duration of the illness. Efforts are needed to develop accurate diagnostic techniques that provide objective information about the health status of the mammary gland during milking and interpretation of results without the need to send samples to laboratories. This can improve control and prophylaxis measures. The objective of this project is to expand information on the use of predictive techniques for the diagnosis and control of bovine subclinical mastitis in voluntarily milked cows. Indexes using machine learning will be developed for tropical conditions, in order to collaborate in the control of subclinical mastitis and reduce the unnecessary use of antimicrobials to control the disease. Interpretations based on machine learning can be made available to make mastitis prevention more efficient and information can be made available for decision-making on rural properties.