Production performance predicting model for the Egyptian dairy buffalo using deep learning

Background In smart livestock farming, machine learning (ML) has shown promising potential for enhancing precision, efficiency and productivity. Aim(s) This study aimed to use a smart production prediction model that can improve the management efficiency of dairy buffalo. Methods Deep neural networks (DNN) were applied depending on eight different recordings (protocol type, sire type, gestation length, lactation length, calving interval, parturition season, open days and dry period) as inputs to predict the calf sex, weight, lactation length, total milk and daily milk yield (DMI), respectively. Two additional traditional ML models, feedforward neural networks (FNNs) and ensemble learning (EL), were also constructed for performance comparison. Major Findings The results showed that DNN, FNN and EL testing accuracy for the calf sex were 86, 83 and 53.4% for lactation length; 78, 70 and 79% for total milk; yield was 78, 68 and 58.9%; and for DMY, it was 82, 70 and 71.2%, respectively. Scientific or Industrial Implications The present model integrates breeding, reproduction and production data to introduce an efficient model for managing buffalo production.