Enhanced GRU-based regression analysis via a diverse strategies whale optimization algorithm

Taking into account the whale optimization algorithm's tendency to get trapped in local optima easily and its slow convergence rate, this paper proposes a diverse strategies whale optimization algorithm (DSWOA) and uses it to optimize the parameters of GRU, thereby achieving better regression prediction effects. First, an innovative t-distribution perturbation is used to perturb the optimal whale to expand the optimization space of the optimal whale. Secondly, in the random search stage, we perform a Cauchy walk on the whale's position and then use reverse learning to enable the algorithm to effectively navigate away from the local optimum. Finally, we adopt a horizontal learning strategy for all whales and use two random whales to determine the current whale's position. Updated, the results suggest that DSWOA is highly effective in global optimization. By utilizing DSWOA, the parameters of GRU were fine-tuned. The experimental findings reveal that GRU produces promising outcomes on multiple datasets, making it a more effective tool for regression prediction tasks.