Skip to content
TAXI NYC ’19, home

Optional

AI settings

Everything on this site works without AI. To try “Ask the data”, paste your own API key. Your browser sends it straight to the provider you pick. It is never sent to this site's server, never logged, and calls are billed to your account.

Provider
AI provider

Default. Cheapest and fastest; $1 / $5 per million input / output tokens.

No key saved for this provider. Create one at platform.claude.com. A key with a low spending limit is a good idea.

AI audit logAI use statement

Conditions

Weather, events and collisions

The 2021 model joined each trip to that day's Central Park weather, the number of permitted street events in its pickup borough and the number of collisions there. Here is every day of 2019 side by side.

Daily series

A year, day by day

Cleaned trips per day

Median trip minutes

Central Park: average temperature (°F, line) and precipitation (in, bars)

Permitted events (bars) and NYPD collisions (line), all boroughs

Day-level relationships

Does the weather move the taxis?

Each dot is a day. Pick a condition and an outcome; the yellow line is a least-squares fit. Rain nudges demand up a little and temperature tracks the seasons, but none of these alone explains much.

Pearson r

-0.39

345 days (1–20 January removed). Weekends in orange. Correlation is not effect: weekday, season and holidays move with the weather.

Hover a dot to read the day.

50K100K150K200K250K300K020406080100Average temperature (°F)Trips that day

The 2021 vendor charts

When trips take longest

Mean trip minutes for every weekday and pickup hour. Weekday afternoons are the slowest; early mornings and weekends are quick. The notebook drew this as one line chart per vendor from a 10% sample; this uses every cleaned trip.

Vendor

Mean minutes from 10.3 (lightest) to 19.1 (darkest)

Mean trip minutes by weekday (rows) and pickup hour (columns)
Weekday12am12 am1 am2 am3am3 am4 am5 am6am6 am7 am8 am9am9 am10 am11 am12pm12 pm1 pm2 pm3pm3 pm4 pm5 pm6pm6 pm7 pm8 pm9pm9 pm10 pm11 pm
Mon12.9 minutes12.2 minutes11.5 minutes11.7 minutes12.7 minutes12.3 minutes12.1 minutes13.4 minutes14.9 minutes15.1 minutes15.1 minutes15.1 minutes15.0 minutes15.1 minutes15.8 minutes16.0 minutes16.3 minutes15.5 minutes13.9 minutes12.5 minutes12.5 minutes12.6 minutes12.5 minutes12.6 minutes
Tue12.6 minutes11.8 minutes11.4 minutes11.5 minutes12.3 minutes11.4 minutes11.3 minutes13.2 minutes15.0 minutes15.8 minutes15.8 minutes15.9 minutes15.9 minutes15.9 minutes16.5 minutes16.7 minutes16.9 minutes16.2 minutes14.5 minutes13.1 minutes12.9 minutes13.0 minutes12.9 minutes12.7 minutes
Wed12.3 minutes11.5 minutes11.0 minutes11.3 minutes12.3 minutes11.5 minutes11.3 minutes13.1 minutes14.9 minutes15.6 minutes15.9 minutes16.2 minutes16.4 minutes16.5 minutes17.3 minutes17.8 minutes18.2 minutes17.5 minutes15.5 minutes13.9 minutes13.5 minutes13.5 minutes13.5 minutes13.1 minutes
Thu12.5 minutes11.7 minutes11.2 minutes11.3 minutes12.2 minutes11.7 minutes11.6 minutes13.2 minutes15.2 minutes16.1 minutes16.4 minutes16.7 minutes16.8 minutes17.1 minutes17.9 minutes18.5 minutes19.1 minutes18.1 minutes16.2 minutes14.6 minutes14.4 minutes14.5 minutes14.4 minutes13.8 minutes
Fri13.1 minutes11.9 minutes11.2 minutes11.3 minutes12.5 minutes12.1 minutes11.9 minutes13.1 minutes14.7 minutes15.5 minutes15.8 minutes16.1 minutes15.9 minutes16.3 minutes17.6 minutes18.2 minutes18.6 minutes17.3 minutes15.3 minutes14.3 minutes14.1 minutes14.1 minutes14.4 minutes14.3 minutes
Sat13.4 minutes12.0 minutes11.3 minutes11.3 minutes12.0 minutes12.8 minutes11.5 minutes10.8 minutes10.8 minutes10.9 minutes11.6 minutes12.4 minutes13.2 minutes13.8 minutes14.5 minutes15.0 minutes15.2 minutes15.4 minutes14.9 minutes14.7 minutes14.5 minutes14.2 minutes14.4 minutes14.2 minutes
Sun13.0 minutes12.1 minutes11.3 minutes11.3 minutes12.0 minutes13.1 minutes12.3 minutes10.9 minutes10.4 minutes10.3 minutes11.0 minutes11.9 minutes12.9 minutes13.9 minutes15.2 minutes15.7 minutes15.7 minutes15.4 minutes14.4 minutes13.7 minutes13.7 minutes13.6 minutes13.1 minutes13.1 minutes

What the model made of it

Eleven numeric features, one survivor

The elastic-net penalty (regParam 0.3, elasticNetParam 0.8) shrank every weather and event coefficient to exactly zero. Only collisions kept a small positive weight, and the trip's zones, hour and rate code did the real work (see the estimator).

Coefficients of the numeric features, minutes per unit
Feature (notebook name)2021 coefficient2026 refit
Precipitation00
Snow00
Snow depth00
TAVG00
WT0100
WT0200
WT0300
WT0600
WT0800
number_of_event00
number_of_collision0.03510.0351