Dynamic Modeling and Analysis of Urban Transportation Networks

This project presents a comprehensive study on predictive modeling of urban transportation dynamics, focusing on the integration and interaction between CitiBike and Yellow Taxi networks in New York City using Graph Neural Networks (GNNs).

Table of Contents

• Project Structure
• Getting Started
  • Prerequisites
  • Installation
• Data Sources
• Usage
• Contributors
• Acknowledgments

Project Structure

project
│
├── 1. Preprocessing
│   ├── Citibike Graph creation function.ipynb
│   ├── dev_preprocess_citibike.ipynb
│   └── dev_preprocess_taxi.ipynb
│
├── 2. NYC Zones
│   ├── Finding centroids of taxi zones.ipynb
│   ├── NYC Taxi Zones.geojson
│   └── ZoneCentroidGeneration.ipynb
│
├── 3. Data Aggregation for time series
│   ├── Daily aggregation of citibike data.ipynb
│   ├── Weekly aggregation of entire citibike data for TGN training.ipynb
│   ├── Monthly aggregation of entire Yellow Taxi data.ipynb
│   └── Monthly aggregation of entire citibike data.ipynb
│
├── 4. EDA
│   ├── sliding_months_EDA
│   │   ├── Sliding months EDA of Citibike.ipynb
│   │   └── Sliding months EDA of Yellow Taxi.ipynb
│   ├── sliding_weeks_EDA
│   │   ├── Sliding weeks EDA Yellow Taxi.ipynb
│   │   └── Sliding weeks EDA citibike.ipynb
│   └── sliding_years_EDA
│       ├── Sliding years EDA of Citibike.ipynb
│       └── Sliding years EDA of Yellow Taxi.ipynb
│
├── 5. gephiGraphs
│   ├── Screenshot 2024-04-17 at 10.35.41PM.png
│   ├── Screenshot 2024-04-17 at 9.52.22 PM.png
│   ├── betweensess.pdf
│   ├── betweensess.png
│   ├── geoLayout13.svg
│   ├── graph1.svg
│   ├── nyc_map.svg
│   └── successful2013.gephi
│
├── 6. Prediction & Report
│   ├── FinalWeightPrediction_WeeklyAggregated-v2.ipynb
│   ├── FinalWeightPrediction_WeeklyAggregated.ipynb
│   └── Report.pdf
│
├──.DS_Store
├── .gitignore
├── requirements_conda.txt
├── requirements_pip.txt
├── README.md

Getting Started

Prerequisites

• Python 3.12 or later
• Conda or pip
• Jupyter Notebook or Jupyter Lab

Installation

Clone the repository:

git clone https://github.com/shyamc757/graph-neural-networks-for-traffic-prediction.git

# Using Conda
conda create --name <env> --file requirements_conda.txt
conda activate <env>

# Using pip
pip install -r requirements_pip.txt

Usage

cd graph-neural-networks-for-traffic-prediction

jupyter notebook

📊 Data Sources

This project uses real-world transportation data from New York City. Please download the following datasets:

🚲 CitiBike NYC Data

• Description: Bike trip records including start/end stations, timestamps, and user types.
• Download: CitiBike System Data

🚕 NYC Yellow Taxi Trip Data

• Description: Trip records for NYC yellow taxis, including pickup/drop-off locations and timestamps.
• Download: NYC TLC Trip Record Data

🌐 NYC Taxi Zone Lookup

• Description: Maps Taxi Zone Location IDs to boroughs and neighborhoods.
• Download: Taxi Zone Lookup Table

💡 Place the downloaded files in a data/ directory, and adjust paths in scripts as needed.

Contributors

• Neel Agarwal (nagarwa9@ur.rochester.edu)
• Shyam Shah (shyamcshah00@gmail.com)

Acknowledgements

• Professor Gonzalo Mateos Buckstein for guidance and insights.
• PyTorch Geometric team for essential tools and libraries.