A Multistep Interval Prediction Method Combining Environmental Variables and Attention Mechanism for Egg Production Rate

The egg production rate is a crucial metric in animal breeding, subject to biological and environmental influences exhibits characteristics of small sample sizes non-linearity. Currently, prediction research predominantly focuses on single-step point prediction, lacking multistep interval exploration. To bridge these gaps, this study proposes recursive, method for rates, integrating variables attention mechanisms. Initially, employed three gradient boosting tree models (XGBoost, LightGBM, CatBoost) the recursive feature elimination (RFE) select critical reduce data dimensionality. Subsequently, by scaling time scale important utilizing variational modal decomposition improved grey wolf optimization (GWO-VMD) time-series decomposition, volume variable augmented its complexity reduced. Applying long short-term memory (LSTM) neural network obtain direct predictions IMFs, predicted outcomes are averaged daily yield upcoming two days. Finally, model based Seq2seq-Attention Gaussian distribution proposed study, parameter carried out using multi-objective algorithm (MOGWO). By inputting historical into model, it possible achieve rates. This was applied analyze dataset rates waterfowl. demonstrated feasibility approach combined with guides estimation regulation husbandry.