Support vector regression for traffic volume forecasting and parameter selection

摘要

Support vector regression (SVR) is a promising method of forecasting traffic volumes. It involves a risk function consisting of the empirical error and a regularized term based on the structural risk minimization principle. The work described in this paper, used SVR to forecast traffic volume. Model selection here served as a key factor in forecasting. A spatio-temporal prediction model was used and parameters were selected by minimizing the upper bound of the Leave-One-Out (LOO) using the BFGS Variable Metric Algorithm. Experimental results show that the use of this method in traffic volume forecasting is feasible and that it provides a promising alternative to traffic volume prediction.