Abstract:
The accuracy of travel time data generated by traveler information systems is a topic of interest for transportation engineers. Several evaluations of data quality have been conducted in recent years to measure the accuracy of these systems. In general, the approach has been to quantify errors from the system by measuring the distance between an estimate and a benchmark or ground truth value. A number of distance-based metrics including average error bias, absolute average error, and root mean squared error, have been used. However, because the benchmark values are random values, the errors are also random and are subject to some uncertainty. This paper proposes a novel approach to measuring errors so that the distribution of the errors can be quantified and a statistical hypothesis test can be constructed. The hypothesis test can be used to decide whether the system is operating at the desired level of accuracy.
