Traffic flow prediction based on improved road network clustering combined with PSOWNN

Aiming at the spatial correlation, nonlinearity and stationary and non-stationary characteristics of urban road network traffic flow data, this paper proposes a prediction method based on the improved traffic network data clustering algorithm based on traffic flow data correlation analysis and wavelet neural network algorithm based on traffic flow segmentation weighted fitness function particle swarm optimization algorithm(MC-MPSOWNN), to improve the prediction accuracy of the algorithm. Firstly, the road network clustering algorithm based on traffic flow data correlation analysis is used to screen out other observation points with high correlation coefficient between the traffic and the predicted traffic flow data, so as to simplify the sample input data and reduce the accuracy of the redundant data. Interference, improve overall prediction accuracy.Secondly, a new particle swarm algorithm fitness function is constructed to give greater adjustment of non-stationary data segments in the overall prediction sample, so as to further improve the prediction accuracy of non-stationary data segments. Finally, the experimental results show that the improved prediction algorithm proposed in this paper improves the convergence speed of the algorithm compared with the pre-predictive algorithm. It also improves the prediction accuracy of the whole and non-stationary data segments and achieves better prediction results.