Non-parametric regression and application for short-term traffic flow forecasting
Huapu Lu · Journal of Tsinghua University(Science and Technology) · 2009
A non-parametric regression technique was developed to improve the accuracy of short-term traffic flow forecasts.The forecasts use trends to dynamically select all input variables with error feedback to tune the system parameters and the input variables.The method has been used for traffic management of a large-scale highway network in Beijing.The results show that the 5 min interval forecasting results from the non-parametric regression are significantly better than neural network results.The feedback to modify the parameters gives better predictions than with unchanged input variables.