This repository contains SpatialDeform, a Python library for turning cost-weighted GeoPandas line networks into travel-time cartograms.
The installable project lives in spatialdeform/ so packaging, documentation,
examples, and tests stay together.
The reproducible NYC experiment lives separately in
experiments/nyc_taxi_2016/. It maps January–June 2016 pickup/drop-off
coordinates directly to the full Manhattan OSM drive graph, fits 168 sparse
full-network cartograms, and compares raw, RMS-normalised and Procrustes PCA
representations. Taxi zones are not part of the active pipeline.
Run the selected model end to end from the repository root:
uv run --project spatialdeform python -m \
experiments.nyc_taxi_2016.run_experiment all \
--input /path/to/train.csv