MMorang-esristaff

Add GTFS to a Network Dataset version 0.5.0.0 released

Discussion created by MMorang-esristaff Employee on Jan 20, 2016

Hello to those of you who are using the Add GTFS to a Network Dataset tool/code sample.

 

Version 0.5.0.0 has been released.  Download it here: http://www.arcgis.com/home/item.html?id=0fa52a75d9ba4abcad6b88bb6285fae1.  Please remember to unregister the previous version of the transit evaluator before registering this one.  Also, always remember to unregister it before uninstalling ArcGIS, or else you might have problems with your registry getting confused.

 

This tool is now also available on GitHub if you want to play with the source code: https://github.com/Esri/public-transit-tools

 

I highly recommend that everyone using this tool upgrade immediately to version 0.5.0.0, as this update corrects a problem that caused incorrect results.  This problem was identified by a user of Add GTFS who ran a large number of OD Cost Matrix calculations.  She found that in a small number of cases (<0.1%), the results were incorrect.  I will attempt to explain the problem and the solution here for those who are interested.

 

Suppose you use the transit evaluator to calculate the travel time between Point A and Point B.  Traveler 0 starts at 10:00 AM, it takes him 1 hour to get to Point B.  Traveler 1 starts at 10:01 AM from the same location, and it takes 59 minutes.  Traveler 2 starts at 10:02 AM, it takes 58 minutes.  This makes sense.  All three travelers end up on the same bus, so they all reach Point B at the same time, even though they started at different times.  Traveler 0 just had to wait one minute longer than Traveler 1 and two minutes longer than Traveler 2.

 

The problem was that in a very small number of cases, Traveler 1 would miss the bus, and it would take, say, 2 hours to reach Point B:

10:00 – 1 hour

10:01 – 2 hours

10:02 – 58 minutes

This clearly does not make sense because Traveler 1 could just wait until 10:02 and reach Point B in 59 minutes instead of starting at 10:01 and taking two hours.

 

The reason for these anomalous results turned out to be a very deep down small rounding error, and version 0.5.0.0 corrects the problem by avoiding these rounding errors.

 

As always, please let me know if you find any problems, and let me know what you’re working on!

 

Thanks!  Happy analysis!

Outcomes