IWOA-RNN: An improved whale optimization algorithm with recurrent neural networks for traffic flow prediction

Abstract

In order to improve the accuracy of traffic flow prediction and support the dynamic control of road management departments, this paper aims to minimize real-time traffic flow prediction errors. Therefore, highway data was cleaned and normalized, and proportionally divided into training and testing datasets. Subsequently, a traffic flow prediction model based on IWOA-RNN is proposed, which integrates the Whale Optimization Algorithm (WOA) with Recurrent Neural Networks (RNN). To address issues such as slow convergence speed and susceptibility to local optima in the WOA algorithm, a population initialization method based on improved Tent mapping is introduced to enhance the diversity of the initial population, thereby improving the global search capability of the algorithm. Simultaneously, a dimension-wise pinhole imaging-based reverse learning strategy is employed to further enhance the local search ability and robustness of the WOA algorithm. These improvements allow more efficient optimization of RNN parameters, improving the performance and accuracy of the traffic flow prediction model. Experimental results on traffic flow data from Guangzhou highways demonstrate that: 1) The improved model significantly outperforms traditional methods in real traffic datasets; 2) It improves the accuracy and robustness of traffic flow prediction, providing new directions and insights for the further development of traffic flow prediction models.

Keywords

• Traffic flow prediction
• Whale optimization algorithm
• Recurrent Neural Network
• Tent chaotic mapping
• Pinhole imaging reverse learning