Travel time predictions: should one model speeds or travel times?

Abstract

Purpose: Travel time predictions are of importance for individual trip planning as well as for logistics applications. Since travel time and travel speed have a one-one correspondence, the modeller has the choice to model travel times directly or model the corresponding travel speeds and infer the associated time from the speed predictions. A priori it is not clear which of these is the superior approach. In this paper we investigate the implications of the choice of the methodology for the accuracy of the travel time predictions. Methods: For a selection of links, travel time prediction models, both in a direct way as well as indirectly via the implied link travel speeds, are obtained. The respective predictions are compared on a validation data set with respect to their accuracy as measured by mean error, root mean square error, mean percentage error as well as mean absolute percentage error. Additionally, the accuracy of route travel time predictions is evaluated based on the raw GPS data from the floating taxis. Results: The empirical results overwhelmingly make the case for using direct modelling if the goal of prediction is to obtain a RMSE-optimal prediction. If the MAPE is to be minimized, however, the indirect method provides the better results. Conclusion: Thus the goal of the prediction determines the better method of modelling: if one is interested in minimizing the RMSE, then, for the data investigated in this paper, the direct method should be selected. However, if one is interested in obtaining a small MAPE, the indirect method achieves better results.

Modelling travel time uncertainty in urban networks based on floating taxi data

“…19% of all measurements are based on only one taxi observation. For more details on the data set see [1]. Figure 2 provides information on the typical travel time measurements: (a) provides the empirical cumulative distribution function (ECDF) of all local travel time obser-…”

“…This information will be In the subfigures the links are colored according to percentage of available (non missing) observations (using the 'hot' colormap of MATLAB; white=0%, orange = 50%, black = 100%). Source: reprinted from [1] used in order to estimate the measurement error variance in Section 3.2.…”

“…On these days for a total of 8 heavily used routes in the four data sets single trip start and end points are estimated. Details on the routes are given in Table 2 of [1], the location of the routes is presented in Fig. 7 in [1].…”

“…Details on the routes are given in Table 2 of [1], the location of the routes is presented in Fig. 7 in [1]. The thus measured travel times are classified into timeof-day-intervals using the estimated time of entry to the route.…”

“…Both private individuals as well as logistics companies typically have asymmetric costs with being late implying higher penalties than being early. Accordingly for cost optimal decisions not only the predicted travel *Correspondence: Dietmar.Bauer@uni-bielefeld.de 1 Bielefeld University, Universitätsstrasse 25, 33619 Bielefeld, Germany Full list of author information is available at the end of the article time but also the uncertainty involved in the prediction is of interest.…”