Sparse Kernel Regression for Traffic Flow Forecasting

In this paper, a new kernel regression algorithm with sparse distance metric is proposed and applied to the traffic flow forecasting. The sparse kernel regression model is established by enforcing a mixed (2,1)-norm regularization over the metric matrix. It learns a mahalanobis metric by a gradient descent procedure, which can simultaneously remove noise in data and lead to a low-rank metric matrix. The new model is applied to forecast short-term traffic flows to verify its effectiveness. Experiments on real data of urban vehicular traffic flows are performed. Comparisons with two related kernel regression algorithms under three criterions show that the proposed algorithm is more effective for short-term traffic flow forecasting.