To run the visualizer, install the pip dependencies and run python app/app.py.
- Using trip records from Jan 2016 through June 2016, available to download from the NYC Transportation and Limousine Commission (TLC) website.
- process_data.py: Several small functions run via the command line to clean and modify data.
- get_zones.py: For each trip record, indicate the TLC zone of the pickup and drop-off locations.
- Data exploration.ipynb: exploratory notebook on the taxi trip records; implementations of most of the Exploration section of my paper are here.
- get_dest_counts.py: Build a csv file where for each hour, day of week, and pickup zone, counts the drop-offs in each drop-off zone. Output used in app.
- get_flow.py: Build a csv file where for each hour and day of week, counts the net taxi trips (drop-offs - pickups) in each zone. Output used in app.
- Regression.ipynb: Using Random forests and SGD regression to predict a trip's drop-off location given its pickup datetime and location; primarily for playing around with scikit-learn.
- grid_search.py: Grid search and cross-validation functions for random forests and SGD; implementation of results in the Prediction section of my paper are here.
- globes.py: Useful variable declarations and function definitions used in other files.
- Network Analysis.ipynb: Turn taxi trips into edges and neighborhoods into nodes; not used in paper.
- zones_*.geojson: GeoJSON zone definitions.
Install the python dependencies with pip install -r requirements.txt. The shapely package requires the GEOS library, which can be installed via brew, apt-get, yum, or the official website.