#A Look at NYC Taxi Data and How We Tip

Computing w/ Large Data Sets - NYU Courant Institute - Fall 2014

Project By:

• Lin, Emily
• Liu, Thomas
• Shaw Susanto, Billy

Main Tools

• Python (main language)
• iPython Notebook (main platform)
• GeoPy (for reverse geocoding)
• Pandas (for data analysis)
• Tableau (for data visualization)

Data

This project uses data by Chris Whong, who has requested all 2013 NYC taxi rides data from NYC Taxi & Limousine Commission. Bless his heart.

We took all trip information, as well as all the fare rides.

We sampled about 75 million rides (>= 30GB) to make the following analysis.

Data Analysis

Basic Information (1 slide per point)

• Which neighborhood takes the cab the most (pickup): 1) List of top 10 2) Pie Chart - "top neighborhood.csv"
• Which is the most popular neighborhood for cab rides (dropoff) 1) List of top 10 2) Pie - chart - "top_drop_neighborhood.csv"
• At which time do New Yorkers take cab ride the most? 1) List of percentage 2) Bar chart - "count_by_hour.csv"
• At which time do New Yorkers take cab rides the least? - "count_by_hour.csv"
• Which time has the lowest avg speed (distance/ time travelled) - "time_speed.csv"
• Which time has the highest avg speed (distance/ time travelled) - "time_speed.csv"

Payments & Tip

• Average in New York CRD - 0.20200465239853405

Specific areas of interest:

• Tip percentage (avg, top)
• Credit card vs. cash
• Total fare

Factors to analyze:

• Average
• Neighborhood: does pick up area/ drop off area affect tip or payment methods (for the top list: only pick with more than 10,000 rides)
• Time of the day
• Distance travelled

Analytics

• Prediction: predicting tip based on neighborhood and time -> lm, simplified lasso. compare two methods by comparing mse.