Track Citi Bike reliability using real-time data

Use this repo to get data from Citi Bike's GBFS real-time data feed and revisit the analysis we used for Riding Forward: Overhauling Citi Bike’s Contract for Better, More Equitable Service

For an overview of this project, see the presentation slides from the 2024 NYC Open Data Week School of Data.

Get data

You will need to start recording Citi Bike's real-time data feed. This template will set up an automatic data getter that will run on GitHub Actions and store data to your own copy of this repo

1. Fork this repo

1. Click 'Fork' above

2. optionally, give it a new name

3. Click 'Create fork'

2. Start recording data

The GitHub Action will fetch and save the GBFS data on the specified schedule. The schedule or frequency is set with crontab syntax. E.g., to fetch data each hour at 10 minutes after the hour, use 10 * * * *

1. set schedule to get data:

• Navigate to /.github/workflows/pull-data.yaml and edit line 4 to specify the frequency to fetch the data.

• replace delete with e.g.

  schedule:
  - cron: '10,25,40,55 * * * *'

  (or whatever frequency/schedule you like)

• Commit the change (and push to GitHub if necessary).

2. enable Actions

• Click the 'Action' tab

• Click I understand my workflows, go ahead and enable them to enable workflows from the forked repo

• on the left, under 'Workflows' click 'pull full data w python'

• Click Enable workflow to enable the workflow

Now data will be saved in the GitHub repo at /data and successful data flow runs will be logged on this page.

3. Build a dataset

After you have recorded some data:

Clone your fork of this repo, with the data now stored in GitHub, to a local instance. You can do this with GitHub Desktop, VS Code, the git command line, or any other way to clone a repo.

Then run the Jupyter notebook Build dataset. This will combine the data snapshots into a single dataset, indexed by last_updated data timestamp and station_id, and a companion geojson of station locations indexed by station_id.

From here, you can use these datasets for any sort of analysis!

Example notebooks

You can follow the steps of the analysis we did through the sample notebooks in Examples.

You can run these on Binder: or clone/download them and run them locally.

1. Build service measures creates the measures of poor service which we used, including frequency and duration stations had no bikes or no docks, and portion of docks which had broken bikes.
2. Compute clusters of poor service identifies significant clusters of poor service across each of these service measures, using local indicators of spatial association.
3. Get Census data and Compare demographics load Census geometry and counts through APIs, join Tracts with the Citi Bike service area and clusters of poor service, and aggregates and visualizes the comparison demographics.
4. Count violations of rebalancing standards estimates the instances and total fine amount for Citi Bike's failure to meet its rebalancing standards requiring stations not remain empty of full.

The sample-data branch of this repo which is already loaded with some data from March 2024. You can switch to that branch to start exploring the analysis over real data.

Python environment

You can create an environment using requirements.txt with Python version 3.9 to load the Python packages needed to run these notebooks.

create a conda environment using the environment.yml to load the Python packages needed to run these notebooks.

Contributing

Please use this template and examples to begin exploring Citi Bike, or use its components for any other analysis!

If you find anything interesting, please reach out to let us know! You can get in touch with Dan directly at dlevine at comptroller.nyc.gov.

Or, submit a pull request with your own work to add to the Examples section here.