Effect of Time Intervals on Short-Term Traffic Flow Prediction Performance for an Adaptive Fuzzy Inference System Approach

Time interval plays a pivotal role for short-term traffic flow prediction since time interval determines traffic data characteristics and, hence, the performances of such prediction methods. In this sense, the effects of time intervals on prediction performance should be investigated systematically when developing short-term traffic flow prediction methods. However, the investigation into time interval effects on short-term traffic flow prediction methods is still limited in the literature, hindering the incorporation of such prediction methods into real-world transportation applications. To this end, this paper proposes an online adaptive fuzzy inference system (FIS)-based short-term traffic flow prediction method with recursive parameter adjustment and designs an experiment to analyze the effect of time intervals on the prediction performance of this proposed method. Using real-world traffic flow data collected from highway systems of the United Kingdom and the United States, the proposed FIS-based method was calibrated through sensitivity analysis, and the effects of time intervals on the proposed FIS-based method were investigated through aggregated performance, group performance, disaggregated prediction performance, and comparative performance. Empirical results showed that the prediction performance of the proposed FIS-based method increases sharply for time intervals from 1- to 10-min, remains stable for time intervals between 10- and 20-min, and decreases slightly for time intervals from 20- to 30-min. This finding delineates the applicable range of time intervals for the proposed FIS-based method, which will be helpful for integrating the proposed prediction method into real-world proactive transportation management systems based on these time intervals. Future studies are recommended to advance the investigation into the effects of time intervals in transportation-related studies.