The Simulation Research of Non-parametric Regression for Short-Term Traffic Flow Forecasting

Short-term traffic flow forecasting play an important pole in urban traffic control and induction system. Non-Parametric Regression (NPR) is one perfect method for short-term traffic flow forecasting based on pattern recognition. At present time, the application research of NPR for short-term traffic flow forecasting is confined in small-scale fields and has less study of forecasting mechanism. This paper is trying to use the simulation measure to research the applicability of NPR in short-term traffic flow forecasting and study the forecasting principles by adjusting different system parameters. A typical road network is constructed as the study object in this paper. The forecasting problems of NPR are studied based on it. The data pre-processing of principal component analysis and cluster analysis are used for better forecasting results. Other network structures have only different simulation parameters with the presented network.