Short-term Traffic Flow Forecasting Model Based On SVR Optimized By Swarm Intelligence Algorithm

Intelligent Transportation Systems are considered as one of the most significant methods to relieve traffic congestion, decrease traffic accidents and improve traffic operational efficiency. Real-time, accurate and reliable prediction of short-term traffic flow forecasting is the key content of realizing intelligent traffic control and guidance, which has great theoretical and practical value.In this dissertation, concept of short-term traffic flow forecasting is introduced as topic, then researches on short-term traffic flow forecasting up to now are summarized. On the basis of learning traffic flow data analysis and support vector regression (SVR) theory, parameters selection of short-term traffic flow forecasting model based on SVR is discussed and studied. Swarm intelligence optimization methods are used to select the optimal parameters, and then simulation results of real data show the effectiveness of the proposed forecasting models. The following are the main conclusions of this dissertation:1. Short-term traffic flow forecasting model based on SVR with Artificial Fish Swarm (AFS) algorithm. For the important influence of forecasting model’s prediction accuracy on the SVR’s parameters selection with punishment coefficient, insensitive loss coefficient and kernel parameter, an Artificial Fish Swarm algorithm is used to optimize the SVR to select the optimal parameters. Along with the adaptive search mechanism of visual and moving step in artificial fish swarm algorithm, a short-term traffic flow forecasting model based on AFS-SVR is proposed. Simulation experiments of real data and results of compared models show that the proposed forecasting model’s feasibility and effectiveness.2. Short-term traffic flow forecasting model based on SVR with hybrid Particle Swarm Optimization (PSO) and Artificial Fish Swarm algorithm. On the basis of AFS-SVR model, for the disadvantage of more parameters and moving step factor’s influence in Artificial Fish Swarm algorithm, a hybrid particle swarm optimization and artificial fish swarm intelligent method is proposed to optimize SVR. This method uses particle swarm optimization formulation to reformulate artificial fish swarm algorithm and chaotic initialization of AFS location information, minimizing the impact of the step factor of artificial fish swarm, which selects the SVR’s optimal parameters. Then a short-term traffic flow forecasting model which combined the SVR model with chaotic hybrid Particle Swarm Optimization and Artificial Fish Swarm (CPSOAFS-SVR) is proposed. Simulation results show that the proposed method yields more prediction performance than single Particle Swarm Optimization and Artificial Fish Swarm optimization model.