Traffic state prediction based on competence region

Abstract

Traffic state prediction is a key basis of traffic flow guidance system and traffic information publishing system. This paper presents a new method of forecasting the traffic state of unban expressway based on competence region. This method can predict the traffic state grade of road based on the distance between the sample data and the traffic state cluster center by creating a competence region of neural network classifier. And this method can effectively integrate the temporal and spatial features together without considering the correlation between the different features, and thus it has a strong adaptability. The experimental results show that this traffic state prediction method can reduce the prediction error and improve the equality coefficients compared with the classical algorithms. The prediction method used in this paper is effective and accurate for forecasting traffic state based on the competence region.