Data Fusion Method for Accuracy Evaluation of Travel Time Forecast

Abstract:

A BP neural network model was brought forward, which was composed by the initial data generated module, the BP network based data fusion module and the result analysis module. Four variables such as link average density, traffic volume, link average travel time based on floating car data(FCD) and floating car sampling size were taken as input variables. Link average density and traffic volume could be obtained by the data of loop detectors, while link average travel time and floating car sampling size could be acquired with FCD. Then, the reasons to choose those four variables were given with the support of a statistical analysis. At last, an arterial road in Hangzhou was chosen as an object link, 406 groups of data were utilized to verify the model. The results show that the mean absolute error (MAE) of the proposed model is only 4.86%.