Based On Chaos And Improve Traffic Flow Forecasting LSSVM Of

The informatization and intelligent construction of transport system is an important measure to accelerate the process of urban modernization, and it is also a basic way to meet the increasing travel demand of urban citizens. Traffic control and guidance is not only the research content of urban transportation planning, but also is an important subject in the field of intelligent transportation, and its implementation depends on the real-time and accurate forecasting of short-term traffic flow. Traffic flow has the characteristics of uncertainty, randomness and time-variation, while chaotic characteristic exists in short-term traffic flow time series. Therefore, how to establish the accurate forecasting model of short-term traffic flow is a hot spot in current research.Based on the collected traffic flow data of a certain road in Qingdao, the paper adopt chaos theory to analyze the chaotic characteristic in short-term traffic flow, so as to lay a foundation for the following forecasting. Firstly, we use the C-C method to dispose traffic flow time series, and obtain the needed embedding dimension and delay time for phase space reconstruction. Then we reconstruct the phase space of traffic flow time series. Finally, the small data method is introduced to calculate the maximum Lyapunov index of phase space to determine the existence of chaotic characteristic.Lease squares support vector machine (LSSVM) has strong learning and generalization capability, and it can dispose small sample and nonlinear problems. To improve the forecasting precision of short-term traffic flow, the paper proposes the forecasting method based on improved LSSVM. We use particle swarm optimization (PSO) to optimize LSSVM, in which the optimal penalty factor and kernel parameter of LSSVM will selected by PSO with global search capability, so the blindness choice of parameters will be avoided. Then the reconstructed phase space will be treated as the input sample of the optimized LSSVM model for forecasting. Experiment results indicate that chaotic characteristic exists in short-term traffic flow, and the forecasting model based on chaos and improved LSSVM not only has better forecasting capability, but also is effective and feasible in the forecasting of short-term traffic flow.