Predicting taxi demand in NYC

Taxi demand prediction in large cities contributes to the satisfaction of a wide audience. The taxi fleet will be working more effectively, the customers will get their rides on time, and even city planners can benefit from knowing the taxi rush hours throughout the day and week.

We demonstrate the taxi demand prediction use case with the NYC taxi dataset (available at https://www1.nyc.gov/site/tlc/about/tlc-trip-record-data.page) provided by the NYC Taxi and Limousine Commission (TLC). The data contains altogether over 1 billion taxi trips from January 2009 forward. For each taxi trip, detailed information is recorded, such as the pick-up/drop-off date, time, and location, the type of the taxi (yellow/green/for-hire vehicle), passenger count, and fare.

The demand prediction means concretely predicting the expected count of taxi trips in the next hour based on the trip counts in the previous hours. The goal is to train the model on data from one...