Citi Bike (NYC)

Fetches trip data straight from the public S3 bucket via bike_network_traffic.get_bike_data and links a Celldega clustergram to the deck.gl flow map.

from bike_network_traffic import silence_warnings
silence_warnings()

import celldega as dega
from ipywidgets import HBox, Layout

from bike_network_traffic import (
    get_bike_data,
    link_flow_to_clustergram,
    make_flow_widget,
    make_station_clustergram,
)

1. Download trips and build station + transition tables

First call downloads (and caches under ~/.cache/bike_network_traffic) the monthly zip from S3 and returns a station table plus a destination-probability matrix ready for Celldega.

ds = get_bike_data('nyc', year=2026, month=3, return_trips=True)
stations, transition_prob, trips = ds.stations, ds.transition_prob, ds.trips
print('stations:', stations.shape, ' transition_prob:', transition_prob.shape, ' trips:', trips.shape)
stations.head()

stations: (2314, 4)  transition_prob: (2269, 2230)  trips: (2945753, 8)

	station_id	station_name	lat	lng
0	7522.02	1 Ave & E 110 St	40.792327	-73.938300
1	7596.11	1 Ave & E 118 St	40.797459	-73.934499
2	5779.08	1 Ave & E 16 St	40.732219	-73.981656
3	5854.09	1 Ave & E 18 St	40.733812	-73.980544
4	6079.03	1 Ave & E 30 St	40.741444	-73.975361

2. Cluster the transition matrix

mat, cgm, cluster_map = make_station_clustergram(transition_prob, n_clusters=150)

3. Build the flow map and link it to the clustergram

flow = make_flow_widget(stations, transition_prob, cluster_map, trips=trips, height=700, debug=True)
link_flow_to_clustergram(flow, cgm)

flow.layout = Layout(width='560px', height='700px')
cgm.layout  = Layout(width='720px', height='700px')
HBox([flow, cgm])