Routing/Travel Time Analysis/Analyzer/Sample Shetland
Shetland
Dataset contains around 830000 trackpoints, sampled at 1Hz frequency. Data were gathered during 5 months long period. Data might be a bit slow-biased because of the used car purpose.
Processing the data I got 7200 travel times for 391 road segments (Shetland map has 980 road segments). 300 road segments had data for both directions, 91 segments only for one direction. At most, there were 59 travel times for one segment in one direction, 180 road segment / direction pairs had only 1 travel time.
Analyzer was able to extract models for 365 road segment / direction pairs, 260 of them had traffic delay details.
The output file with models can be downloaded from SVN. You can get a brief overview of the results from the animation at [1].
Results were tested on 10 random tracks (different tracks, different times of the day)
track # | real travel time [s] | estimated travel time [s] | error [%] |
---|---|---|---|
1 | 618 | 617 | 0.0 |
2 | 875 | 830 | 5.1 |
3 | 1733 | 1490 | 13.8 |
4 | 345 | 321 | 6.9 |
5 | 1589 | 1215 | 23.3 |
6 | 454 | 474 | 4.4 |
7 | 873 | 762 | 12.7 |
8 | 1843 | 1698 | 7.8 |
9 | 238 | 215 | 9.6 |
10 | 1727 | 1530 | 11.6 |
Algorithm usually underestimates travel times. Average error was 9.7%.
Data were provided by J.R. Peterson and data acquisition was sponsored by Cloudmade - Thanks!