Forecasting of Short-Term Traffic Flow Based on SVR with SFLA

Abstract

The forecasting of accurate short-term traffic flow is the key issue in intelligent transportation systems(ITS), and it is also an important prerequisite of real-time traffic signal control, traffic assignment, route guidance, incident detection, ect. In this paper, a SFLA-SVR forecasting model of short-term traffic flow is proposed combining with the support vector regression(SVR) and shuffled frog-leaping algorithm(SFLA), due to the features of nonlinearity, complexity and randomicity of short-term traffic flow. Support vector regression(SVR) has been successfully employed to solve regression problem of nonlinearity and small sample, but it is very crucial to select appropriate parameters of learning accuracy and generalization performance of SVR, then, shuffled frog-leaping algorithm(SFLA) is used to determine free parameters of support vector regression. Two adjacent intersections of the typical urban road network are selected for the study object, the experimental results demonstrate that SFLA-SVR outperforms the BP neural-network model in prediction accuracy.