With the improvement of the requirements for intensive breeding of calves in large-scale pastures in China, in terms of nutrition, immunity and management efficiency, a feeding information system was designed to realize the visual management of data in the whole feeding process of calves of different ages，aiming to adapt to the needs of the reform of the working mode of breeding managers. Improved Logistic regression algorithm was used to predict the milk supply of calves to realize automatic and accurate feeding, and ensure the healthy development of calves. Based on the B/S architecture, a calf feeding information management system was designed, which can collect the basic information of calves, calf weight, historical milk drinking information of calves, real-time milk drinking information of calves, and working status of milk drinking stations, and all kinds of data of the elements, realizing the front-end display and back-end storage of the data, and visualization of the data in the whole process of calf feeding. Based on the Logistic regression algorithm, a prediction model for the milk supply and milk replacer concentration of calves was established，the running time of the improved algorithm can be shortened to 0.3s, and the efficiency of the algorithm can be increased by 12 times. The calf breeding test showed that the prediction model had good accuracy. The average actual milk drinking rate of calves in the experiment reached more than 95%. The intelligent and refined level of cattle management reduced feeding costs.