Abstract: Because wavelet is suitable for processing nonlinear,random signals and support vector machines excel at solving less-data,nonlinear,multi-dimension problems,the paper proposes combining of wavelet package with least squares support vector machines for short-term traffic flow forecasting. First,theories of wavelet package and least squares support vector machines are introduced,and then a short-term traffic flow forecasting method based on wavelet package and least squares support vector machines is proposed. Second,the effect of the method is tested by the real-time traffic flows collected in Beijing City. The result shows the feasibility and validity of the proposed method.

Key words: short-term traffic flow forecasting, wavelet package, support vector machines, stab stical learning