An approximation of service frequency

October 1st, 2013

Here is a rather crude, though I think useful, visualization of service frequency at the stop level. Basically, I used the GTFS data from SORTA and TANK to calculate the number of times a bus stops at each stop every week. Since a week is the basic cycle period of transit(service is bad on Sunday, better on monday), this should give us a an idea of basic average frequency with the huge caveat that there’s enormous variation within each week.

Click the image to get a bigger version. There’s lot’s of interesting detail in there!

You may notice that frequency can appear vary in a single line where it doesn’t seem like it probably should:

Ludlow Avenue

In most cases, this is simply an artifact of the way I grouped stops that were next to each other and had exactly the same name. At least 2-3,000 stops of the 6,000 stops in the dataset can reasonably be thought of as pairs with one serving each direction of travel.

2 responses to “An approximation of service frequency”

1. Tim Bender says:

This data is just face-value, I know. More dot density doesn’t mean better transit service, it just means more stops – which sometimes can be a bad thing.

NKU is a hot spot because of the NKU Shuttle, a route that is very localized and very frequent but isn’t intended to provide connectivity to the greater network. I’m surprised however that this isn’t more apparent in the visualization.

Ritte’s Corner in Latonia (Southern-Decoursey-Winston intersection on the #7) is a hot spot, most likely because every #7 bus passes that intersection twice per one-way trip.

Nice exercise!

• Nate Wessel says:

An important point, regarding dot density! I’m currently working on a linear version of this that should give us a much more accurate picture, though it’s going to be much more complex to create if I’m able to do it at all. I’m trying to incorporate degrees of schedule (un)coordination as well.

Since the circles are scaled by area, some of the outliers like the one at Ritte’s Corner blend in more than you might expect. A stop with a value of 10 for example would get represented by a circle with a diameter of (√(10/pi))*2 = 3.568 while a stop with a value of 100 would get only (√(100/pi))*2 = 11.284.

Maybe coloration for the outliers would have added something to the visual though it probably would have just made a useless cluster downtown.