Abstract: By integrating variational mode decomposition(VMD), the dung beetle optimization(DBO)and the bidirectional gated recurrent unit(BiGRU), a price prediction model(VMD-DBO-BiGRU)is proposed, aiming to enhance iron ore futures prediction accuracy. First, VMD is employed to decompose the iron ore futures price series into different modal components, extracting multi-scale price features while suppressing noise. Then, DBO is utilized to optimize the hyperparameters of the BiGRU model, mitigating the risk of local optimum associated with traditional optimization methods. Finally, the optimized BiGRU model is applied to predict each decomposed component, and the final futures price prediction is obtained by linear reconstruction of the predicted results. Experimental findings demonstrate that the proposed model significantly improves prediction accuracy in both single-step and multi-step forecasting, consistently outperforming benchmark models. Moreover, this model provides robust technical support for hedging and investment decision-making in enterprises such as a steel plant in Shandong, contributing to risk mitigation and improved decision efficiency.

Key words: iron ore futures price, variational mode decomposition, dung beetle optimization algorithm, bidirectional gated recurrent unit