Research On Short-term Traffic Flow Prediction Based On Chicken Swarm Algorithm

Intelligent transportation system can alleviate the urban congestion,and the core is short-term traffic flow prediction,so the accurate and rapid short-term traffic flow prediction is of great significance.However,there is random fluctuation and high nonlinearity in short-term traffic flow data,which can't make the prediction speed and precision of many forecasting methods satisfactory at the same time.Wavelet Neural Network(WNN)has good time and frequency domain characteristics,Chicken Swarm Optimization(CSO)has better global convergence and computational robustness,and CSO can find the global optimal solution more quickly compared with Particle Swarm Optimization Algorithm(PSO)and Wolf Pack Algorithm(WPA).So CSO is used to find better weights and wavelet factors for WNN first,then this WNN is utilized to predict short-time traffic flow,for the further improvement of prediction accuracy,CSO is improved in the thesis,finally the Radial Basis Function(RBF)neural network is used to further improve the prediction model.The short-term traffic flow prediction models proposed in this thesis are all tested on open data sets,the innovation points are as follows:(1)The CSO is used to find better weights and wavelet factors for WNN,that is CSO-WNN model.Compared with original WNN,the accuracy of CSO-WNN model is improved by 0.5%,and the speed is enhanced by 9%;compared with WPA-WNN and PSO-WNN model,the prediction speed of CSO-WNN model is improved by 42% and 6.5% respectively.(2)The CSO is improved,then combined with WNN,that is ICSO-WNN model.Compared with CSO-WNN,the accuracy is improved by 0.35%.(3)RBF is used to constitute a combined model ICSO-WNN-RBF,then the model is utilized to predict the linear and nonlinear parts of traffic flow data.Experiments show that the prediction accuracy of ICSO-WNN-RBF is 98.888%,which is enhanced by 0.27% compared to the model based on SVR in the same condition.