Research on Dew Intensity Prediction Based on <scp>ARO</scp> ‐Optimized <scp>LSTM</scp> ‐ <scp>GRU</scp> Hybrid Model

ABSTRACT Dew prediction is of great significance for agricultural soil management, ecological water balance maintenance, and the optimization of meteorological services. Nevertheless, the data obtained from existing prediction methods are characterized by high uncertainty and substantial noise. This situation not only makes it difficult to capture crucial patterns but also presents notable challenges for subsequent research on dew intensity prediction. In this study, a novel dew prediction approach is proposed, which combines an enhanced artificial rabbit optimization (EARO) algorithm with a hybrid long short‐term memory (LSTM)‐gated recurrent unit (GRU) model. To deal with the detour foraging, random hiding, and energy conversion mechanisms in the original ARO algorithm, Levy flight, particle swarm optimization (PSO) guidance, and an adaptive local search mechanism are introduced. These additions effectively enhance the convergence speed and accuracy of the algorithm. The improved algorithm is compared with other benchmark swarm intelligence algorithms. All algorithms are tested and validated using the CEC2022 test functions. The EARO algorithm is then utilized to optimize the hyperparameters of the LSTM‐GRU model, including the learning rate and the size of the hidden layer. Meteorological data from the Fujin swamp wetland in Northeast China are chosen as the research sample. The optimized LSTM‐GRU model is compared with other benchmark models, and performance evaluations are carried out using metrics such as mean squared error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and the coefficient of determination ( R 2 ). The results show that: (1) the data processing workflow significantly improves the stability of the original data; (2) the automatic optimization of the LSTM‐GRU model's hyperparameters by the EARO algorithm minimizes human interference and enhances the prediction accuracy.