Urban Mobility Index
An ML model trained on 600+ neighborhoods predicts walkability from street connectivity and transit density — giving planners a score for any location, even without pre-calculated metrics.
Approach
Walkscore is a great metric — when it exists. For neighborhoods Walkscore.com hasn't pre-computed, planners are stuck. We trained an ML model that predicts a walkability score from street network shape and transit density, so any neighborhood can have one.
Data Pipeline
Three inputs feed the model: road network screenshots from Google Maps (classified with Keras), bus stop locations from the Overpass API, and intersection node counts pulled from OpenStreetMap. We trained on six cities — Boulder, Ann Arbor, Chicago, Washington DC, New York, and San Francisco — and validated on three more: Madison, Seattle, and Tulsa.
The Tool
Model Results
We evaluated three clustering algorithms—K-Means, Agglomerative, and Gaussian Mixture—with Gaussian Mixture yielding the most realistic urban clusters. Linear regression using bus stop and intersection densities as predictors achieved a mean Walk Score of 71.07, with RMSE of 17.04 and an R-squared value of 0.38, indicating these features explain 38% of walkability variance.
Limitations & Next Steps
The training cities skew dense and coastal — the model struggles on cities outside that distribution. The R-squared of 0.38 says it: bus stops and intersection density only explain part of what makes a place walkable. To get serious, the next version needs more cities, more input variables (block size, street trees, sidewalk width), and a usable web frontend so a planner can paste an address and get a number back.