Short‐term traffic flow prediction using fuzzy information granulation approach under different time intervals

Abstract

Short-term traffic flow forecasting has been regarded as essential for intelligent transportation systems, including both point prediction and interval prediction. Compared with point prediction, interval prediction of traffic flow in the future will be critical for traffic managers to make reasonable decisions. This study applies the fuzzy information granulation method to obtain the dispersion range of the collected traffic flow time series, and classical forecasting approaches of K -nearest neighbours, back-propagation neural network, and support vector regression are applied on the dispersion range and the original series itself, constituting a short-term traffic flow forecasting system with the capability of joint point and interval prediction. Using real-world traffic flow data collected from a field transportation system in America, the proposed forecasting system is shown to generate workable point prediction and associated prediction interval, demonstrating the validity of the proposed forecasting system. In addition, for unravelling the impact of time interval on the forecasting system, different time intervals are investigated, showing that with the increase in time interval, the stability of the forecasting system increases. Discussions are provided for the proposed approach, and future works are expected to enhance the proposed forecasting system.