This map displays Citi Bike trips taken in April of 2019. The default layer displays each station in the system, along with its total trips. Clicking on a point will show the number of trips that started at that station, and how many of those trips ended at each station in the system.
See map at https://cgettings.github.io/Citi-Bike-Trips/
TODO:
Click on stations to change focus without having to clear map first- Add other months
- Better handling of touch screens
Fitting a generalized linear mixed effects model to trip, station and weather data, starting from the raw data files.
Trip data can be retrieved from the Citi Bike system data page (or from the AWS S3 Bucket).
downloading_trip_data.R
trips_database.R
Station status data can be retrieved from Citi Bike's GBFS data feed.
building_station_status_database_from_archive.R
station_status_database.R
Data obtained through NOAA NCEI API.
combining_trips_station_weather_data.R
computing_covariates.R
diagnostic_plots.R
resid_patterns.R
computing_model_predictions_on_grid.R
visualizing_model_predictions.R
Example plots are in the folder plots/time_series/2018/glmmTMB_fit_m345_20s_genpois_141213, and are based on a model fitted on data from March, April, and May of 2018 ("m345"), with a sample of 20 stations ("20s"), using a generalized Poisson distribution ("genpois").
Scatterplot heatmap using ggplot2::geom_hex(), with binwidth = 1; original points represent number of trips during a single hour, on a single date, at a single station. Most fitted values are within 7 trips of the observed value, and fitted values are pretty evenly distrbuted around the 1:1 line, with a bias towards under-predicting number of trips, which is to be expected when using a mixed model.
