Research on Traffic Data Prediction Model Based on GJO-GRU

This is the authors' abstract. We don't add key points for this paper.

To address the issue of low prediction accuracy caused by the high randomness of traffic flow, this paper proposes a data prediction model called GJO-GRU. The model utilizes the Golden Jackal Optimization (GJO) algorithm to optimize the Gate Recurrent Unit (GRU) and overcome the effects of nonlinearity and temporal variations. Firstly, to tackle the challenges of high complexity and slow training speed in Long Short-Term Memory (LSTM) neural networks, a simpler GRU structure is adopted as the prediction model. The GJO algorithm is employed to optimize the hyperparameters of the GRU model, reducing the impact of human factors and improving the accuracy of the model predictions. Finally, the proposed GJO-GRU model is compared with GRU, LSTM, and GJO-LSTM models using the publicly available California Performance Measurement System (PeMS) dataset. Experimental results demonstrate that the GJO-GRU neural network prediction model achieves the highest coefficient of determination and the lowest root mean square error (RMSE), mean absolute percentage error (MAPE), and mean absolute error (MAE). This indicates the model's strong fitting and generalization capabilities, significantly enhancing the accuracy of traffic flow prediction.