End-to-end deep SAE-DNN model for predicting Egyptian buffalo calf sex, weight, and daily milk yield

In the present study, a novel stacked Sparse Autoencoder-Deep Neural Network (SAE-DNN) learning prediction model was applied to predict calf sex, weight, and daily milk yield for dairy buffalo. First, SAE stage extracts the unique statistical features of calf records (protocol type, sire, gestation length, lactation season, calving interval, parturition season, open days, and dry period). Next, the DNN stage utilizes the extracted statistical sparse features to predict calf sex, weight, and daily milk yield production. The results showed that the proposed SAE-DNN model introduces a robust model with 80% accuracy, an R-squared value of 0.9149, a Mean Absolute Error (MAE) of 0.2097, and a Root Mean Squared Error (RMSE) of 0.4579 for calf sex. An accuracy of 87%, an R² value of 0.7122, a MAE of 3.9394, and an RMSE of 2.7635 for calf weight. As well as 86% accuracy, R2 value of 0.7964, a MAE of 2.0475, and an RMSE of 2.7798 for daily milk yield.