No abstract

In data-driven travel-time prediction, previous studies have mainly used speed as the input. However, from a traffic engineering perspective, given that speed varies little in the free-flow regime, traffic density, which can accurately represent traffic conditions from the free-flow regime to the congested-flow regime, is preferable as an input. In this study, we compared the accuracy of traffic densities spatially interpolated using spatial statistical and machine learning methods, and validated their effectiveness as inputs for travel-time prediction. The results show that even traffic density interpolated by simple spatial interpolation contributes to the accuracy of travel-time prediction and is superior to speed for early detection of traffic congestion.

No abstract

No abstract