Feature selection to traffic flow forecasting

Traffic flow forecasting is a popular research topic of intelligent transport systems (ITS). Some efficient forecasting models have been developed. In practice, many feature variables are available for traffic flow forecasting. It is important that how to choose the most informative forecasting variable combination, which can save lots of computation cost of forecast and improve forecasting precision. In this paper, mutual information is used as a metric to measure the correlation between traffic flow variables. Feature selection based on mutual information is generalized to regression problems and is used to select the most informative variable combination. Nonlinear regression Support Vector Machines (SVM) network is used as the forecasting model. Bayesian inference is used to determine the kernel parameters of the SVM. The efficiency of the method is illustrated through analyzing the traffic data of Jinan transport system.