Objective To address the significant fluctuations in temperature and humidity within enclosed pig barns caused by the lag in environmental control equipment regulation, an appropriate multi-element time series temperature and humidity prediction model is proposed.

Method Pearson correlation analysis was employed to determine the correlation among the 12 environmental factors collected, thereby preliminarily selecting the input features for the model. The selected input features were normalized to eliminate the influence of data scale. DDGCRN, long short-term memory network, support vector regression, and random forest models were chosen. The model prediction results were validated, and the best-performing model was selected.

Result The input features of the temperature and humidity prediction model were determined. Through comparative verification, the DDGCRN model demonstrated the highest prediction accuracy. It exhibited an average absolute error of 0.079 and 0.458 for temperature and humidity predictions respectively, with a root mean square error of 0.134 and 0.719, and an average absolute percentage error of 0.392% and 0.675%. The comparative analysis of model input configurations revealed that an excessive number of input features might not improve the predictive ability of the model and could potentially decrease it. Additionally, different types of models and different prediction targets required different suitable input features.

Conclusion The use of the DDGCRN temperature and humidity prediction model can provide early warning of temperature and humidity changes within the barns, thereby offering valuable insights for precise control of temperature and humidity in breeding environment.