Space vs. Place
In urban analysis, there’s a big difference between space (coordinates, zones, storage) and place (how people actually use and experience a location). A “Restaurant Zone” on a zoning map might not match the bustling, late-night food district that locals actually know.
PlaceCrafter is a new web-based tool designed to bridge this gap. It ignores administrative borders and instead uses functional clustering to reveal the “real” geography of a city based on Point of Interest (POI) density.
PlaceCrafter doesn’t just show you where things are—it shows you how they cluster to form meaningful neighborhoods.
The Workflow
The tool guides researchers through a 4-phase “Platial Discovery” process:
1. Filtering (The “What”)
Users connect directly to the OpenStreetMap Overpass API to pull in live data. But instead of just “all points,” they build semantic profiles—e.g., “Show me a ‘Nightlife’ layer made of pubs, clubs, and late-night takeaways.”
2. Clustering (The “Where”)
PlaceCrafter applies selectable machine learning algorithms directly in the browser to identify hotspots:
- DBSCAN: Finds organic, irregularly shaped clusters based on density.
- K-Means: Forces data into compact, comparable regions.
- Hierarchical: Reveals nested structures (e.g., how a specific bar street sits inside a larger entertainment district).
3. Validation (The “Why”)
It’s not just a visual guess. The system runs real-time statistical tests to confirm that a cluster is “real”:
- Moran’s I: Checks for spatial autocorrelation (basically, “is this clustering random or significant?”).
- Silhouette Scores: Measures how tightly grouped the points are.
4. Visualization (The “readable” City)
Finally, it renders these clusters using novel “fuzzy” visualization techniques. Instead of hard borders, it uses gradient spray cans, convex hulls, and influence grids to represent distinct places as they truly are: distinct cores with fuzzy, overlapping edges.
Case Study: Nottingham
In our test deployment for the city of Nottingham, PlaceCrafter analyzed 534 “Leisure” and “Heritage” POIs. It successfully identified 18 distinct functional regions that crossed official ward boundaries, revealing a “Heritage Corridor” that locals know but that doesn’t exist on any official map.
Future Impact
This work, presented at OSMScience 2025, pushes the boundaries of “Platial Information Systems.” By allowing planners to see the functional structure of a city—not just its administrative one—we can design services, transport, and public spaces that align with how people actually live.