NYU-CUSP-Capstone-2019

• This is our capstone projects of NYU CUSP 2019. The topic is: Does Uber/Lyft reduce parking violations in NYC?
• Group member: Junjie Cai, Junru Lu, Pranay Anchan, Shijia Gu and Yuxuan Wang
• Sponsor & Supervisor: Zhan Guo
• Please refer to our website for more visualizations

Introduction

• This capstone project aims to explore one potential Uber/Lyft’s impact: whether daily Uber/Lyft trips affect parking violations. NYC daily Uber/Lyft trip and parking ticket data are collected and correlated by taxi cab zone. Three technical models, Fixed Effects, Difference in Difference, and Bayesian Network, are applied on the prepared data. The results of these models show the negative correlation and causal effect between the number of Uber/Lyft trips and parking tickets, suggesting Uber/Lyft help in reducing parking violations in NYC. Given the controversial issues around TNC, this capstone project can assisting in understanding impact of Uber/Lyft and offer policy insight to the TNC regulation. (Full report, PPT)

Usage

Step 1: Data Collection

• Data Collection: Developed a pipeline to collect all the 42-month FHV and 6-year parking ticket datasets
• Note: all data is open on the Internet. For every data, we've provided either the link or the intermediated form transformed by us. Also, feel free to change data pathes when re-using these codes, as the data may be called differently in our notebooks

Step 2: Data Preprocessing

• Basic Geo-unit of all data: taxi zone

OBJECTID / LocationID	Shape_Leng	Shape_Area	borough
Taxi zone ID	Taxi zone length	Taxi zone area	NYC borough number

• Uber & Lyft Data: Filtered and grouped the Uber/Lyft trips by taxi zones from FHV trips with Spark.

• Tickets Data

  • Street Name: Filtered and grouped ticket data with Spark
  • Geocoding: Converted 350 thousand street names into coordinates through Google Geocoding API
  • Taxi Zone: Mapped the coordinates into taxi zones with R-tree method with Spark.

• Additional Data

Temporal Data	Spatial Data	Potential Data
Weather, Holiday and Weekdays	ACS, Crime, Transportation, Education and Parking Facilities	Events, BBL, Park, Parking Regulation Locations and Signs, Meter parking price, Garage parking price, Google POI and Yelp

ACS details

DensityPop	IncomePerCap	Poverty	Professional	Service	Office
Population Density	Income per capita ($)	% under poverty level rate	% employed in management, business, science, and arts	% employed in service jobs	% employed in sales and office jobs

Production	Employed	Unemployment	Drive	Carpool	Transit	Walk
% employed in production, transportation, and material movement	% employed rate (16+)	% Unemployment rate	% commuting alone in a car, van, or truck	% carpooling in a car, van, or truck	% commuting on public transportation	% walking to work

OtherTransp	WorkAtHome	MeanCommuteMean	Construction
% commuting via other means	% working at home	commute time (minutes)	% employed in natural resources, construction, and maintenance

Crime details

FELONY	VIOLATION	MISDEMEANOR
Number of felony crimes in the taxi zone	Number of violation crimes in the taxi zone	Number of misdemeanor crimes in the taxi zone

Transportation, Parking Facilities and Education details

subway	bus	meter	parkinglot	sat
Number of subway entrances	Number of bus stops	Number of merter parking	Area of parking lot	Average score of SAT reading, math, and writing

Step 3: Data Integration

• Data Integration:
  • Collected and processed influencing factor datasets, including ACS Census, Crime, Transportation, Education, and Parking data
  • Integrated and output all datasets through spatial/ temporal dimension

Step 4: Data Analytics

• Preliminary Analysis:
  • Time series analysis for the whole NYC and every taxi zone
  • Correlation analysis between parking ticket and FHV trips for the whole NYC
  • Correlation analysis for every taxi zone

Step 5: Modelling

• OLS: Multivariate linear regression modeling trial on temporal datasets.
• Fixed Effect & Bayesian Nework: Fixed Effect Modelling & Bayesian Nework Modelling
• Difference in differences: Mimic the design of experimental research, studying the differential effect of Uber/Lyft on NYC's parking violation