Services on Demand
On-line version ISSN 1678-4596
NAAS, Irenilza de Alencar; QUEIROZ, Marcos Paulo Garcia; MOURA, Daniella Jorge de and BRUNASSI, Leandro dos Anjos. Dairy cows estrus estimation using predictive and quantitative methods. Cienc. Rural [online]. 2008, vol.38, n.8, pp.2383-2387. ISSN 1678-4596. http://dx.doi.org/10.1590/S0103-84782008000800048.
Brazil is the sixth worlds larger milk producer, increasing its production at an annual rate of 4% above other producer countries. Part of this raise in milk production was due to the use of several technologies that have being developed for the sector, mainly those related to genetics and herd management. Accurate estrus detection in dairy cows is a limiting factor in the reproduction efficiency of dairy cattle, and it has been considered the most important deficiency in the field of reproduction. Failing to detect estrus efficiently may cause losses for the producer. Quantitative predictive methods based on historical data and specialist knowledge may allow, from an organized data base, the prediction of estrus pattern with lower error. This research compared the precision of the estrus prediction techniques for freestall confined Holstein dairy cows using quantitative predictive methods, through the interpolation of intermediate points of historical herd data set. A base of rules was formulated and the values of weight for each statement is within the interval of 0 to 1; and these limits were used to generate a function of pertinence fuzzy that had as output the estrus prediction. In the following stage Data mining technique was applied using the parameters of movement rate, milk production, days of lactation and mounting behavior, and a decision tree was built for analyzing the most significant parameters for predicting estrus in dairy cows. The results indicate that the prediction of estrus incidence may be achieved either using the association of cows movement (87%, with estimated error of 4%) or the observation of mounting behavior (78%, with estimated error of 11%).
Keywords : predictive modeling; data mining; Fuzzy logic.