基于IBWO-Stacking的猪舍环境预测模型优化

Optimization of IBWO-Stacking-based Environmental Prediction Model for Swine Housing

摘要:

稳定的温湿度环境对于妊娠母猪的健康和生产性能至关重要。 为解决传统预测模型在妊娠猪舍多因子耦合场景中适应性不足的问题,提出了基于改进白鲸优化算法(Improved beluga whale optimization,IBWO)和 Stacking 集成学习的猪舍环境预测模型 IBWO-Stacking。 在原白鲸优化算法的基础上引入 Logistic 混沌映射、自适应变量的调整以及 Pattern Search 提高 BWO 的寻优能力,实现了对猪舍温湿度预测模型超参数的精准优化,预测猪舍环境中的温湿度。 实验结果显示,IBWO-Stacking 模型预测相对湿度的平均绝对误差、均方误差和均方根误差分别为 1. 123 7% 、3. 243 4% 2 和 1. 800 9% ,预测温度时 3 项误差指标分别为 0. 210 4℃ 、1. 216 5℃ 2 和 1. 102 9℃ ;与 BWO Stacking 模型对比,预测猪舍相对湿度和温度的模型决定系数分别提高了约 0. 55% 和 1. 32% ,平均绝对误差降低约 7. 35% 和 6. 36% 、均方误差降低约 17. 23% 和 30. 92% ,均方根误差降低约 9. 02% 和 16. 9% ,验证了 IBWO-Stacking 模型具有较好的预测性能和鲁棒性,为妊娠猪舍的智能化环境调控提供决策支持,助力减少环境应激对母猪繁殖性能的负面影响。

Abstract:

A stable temperature and humidity environment is crucial for the health and production performance of pregnant sows. To address the inadequacy in adaptability of traditional prediction models under multi-factor coupling scenarios in gestation pig houses, an IBWO-Stacking prediction model was proposed based on improved beluga whale optimization (IBWO)and Stacking ensemble learning. The proposed method enhanced the optimization capability of the original beluga whale optimization (BWO)algorithm through three key improvements: Logistic chaotic mapping initialization, adaptive variable adjustment, and pattern search integration. These enhancements enabled effective hyperparameter optimization of base learners in the Stacking ensemble framework, the model predicted environmental temperature and humidity. Experimental results demonstrated that the IBWO-Stacking model achieved superior performance metrics: for humidity prediction, the mean absolute error (MAE), mean squared error (MSE), and root mean squared error (RMSE)were 1. 123 7% , 3. 243 4% 2 and 1. 800 9% respectively. For temperature prediction, these metrics reached 0. 210 4℃, 1. 216 5℃ 2 and 1. 102 9℃. Comparative analysis with the baseline BWO-stacking model showed significant improvements: the coefficient of determination (R2 )was increased by approximately 0. 55% for humidity and 1. 32% for temperature prediction. Concurrently, MAE was decreased by 7. 35% and 6. 36% , MSE was reduced by 17. 23% and 30. 92% , while RMSE declined by 9. 02% and 16. 9% for humidity and temperature predictions respectively. These results validated the enhanced predictive capability and robustness of the IBWO-Stacking model, providing decision-making support for intelligent environmental control in gestation pig houses, thereby mitigating the negative impacts of environmental stress on the reproductive performance of sows.