GTUP (1)
Description
This dataset accompanies the paper "Graph-Theoretic Urban Planning (GTUP): A Multi-Layer Framework for Cities as Complex Networks," which analyses two morphologically contrasting U.S. cities — Fairfax, Virginia (suburban, cul-de-sac-dominated) and Alexandria, Virginia (historic grid with suburban fringe) — as multi-layer coupled networks integrating transport, land-use/points-of-interest, and green infrastructure. All artefacts are derived from the OpenStreetMap snapshot recorded in the accompanying manifest and are projected to EPSG:32618 (UTM Zone 18N). The archive contains: (i) cleaned simple-undirected WALK and DRIVE street networks for both cities as ESRI shapefiles, with OSM node identifiers, vertex degrees, odd/even parity flags, and metric edge lengths; (ii) Minimum Spanning Trees of each network computed with Kruskal's algorithm; (iii) Chinese Postman matched-edge routes — the shortest-path polylines representing edges that a Eulerian closed walk must traverse twice, obtained by exact min-weight perfect matching on odd-degree vertices with k-nearest-neighbour sparsification verification (k = 30 vs 60, Δ agreement < 10⁻⁴%); (iv) the supplementary HAS* optimal TSP tour over 113 Fairfax POIs, solved via Google OR-tools Guided Local Search; (v) POI features classified into four Moreno-category layers (commerce, education, healthcare, recreation) and green-infrastructure polygons; (vi) node-level CSV tables including betweenness centrality, per-category 15-minute accessibility distances, and supra-adjacency rank comparisons; (vii) per-cell JSON outputs recording every metric with its formula, algorithm, and computation time; (viii) all manuscript figures as PNG; and (ix) a complete reproducibility manifest listing pinned library versions, random seeds, parameter values, and run timestamps.
Files
Steps to reproduce
The pipeline is a Python 3.12 notebook designed to run on Google Colab Pro with a High-RAM CPU runtime (approximately 2 hours wall-clock end-to-end). Fixed pinned versions are required for exact reproducibility: OSMnx 2.0.1, NetworkX 3.3, GeoPandas 1.0.1, Shapely 2.0.6, SciPy 1.13.1, NumPy 1.26.4, pandas 2.2.2, OR-tools 9.10.4067. A fresh Colab runtime requires one kernel restart after the first pip-install block; the provided setup cell handles this automatically. Clone the UrbanGraphPy v2.0 repository and open the accompanying notebook. Mount Google Drive so that the workspace directory Graph_Theory/Phase1 becomes persistent across sessions; OSMnx caches OSM Overpass responses into this directory, preserving the exact snapshot used for this paper. Cell 0 installs pinned dependencies, sets the random seed to 42, and records environment details in manifest.json. Cells 1–8 execute the Fairfax analysis sequentially: data acquisition, simple-undirected graph conversion (the critical step that corrects the MultiDiGraph representation artefact identified in v1), Chinese Postman excess via exact min-weight matching on k-NN-sparsified complete graph of odd-degree vertices, MST redundancy on both WALK and DRIVE networks, betweenness centrality (exact on DRIVE, approximate k = 500 sampling on WALK), per-category 15-minute-accessibility P15M with multi-source Dijkstra, supplementary HAS* with OR-tools guided local search, and multi-layer supra-adjacency betweenness with Spearman and top-20 Jaccard comparison against single-layer rankings. Cell 9 replays the full pipeline on Alexandria, VA, emitting side-by-side comparison tables. Cell 10 writes the cross-city summary figure and master JSON manifest. Cell 11 exports every graph, MST, matched-pair route, and TSP tour as ESRI shapefiles in EPSG:32618, bundles all JSON and CSV results with PNG figures and the manifest, and produces the single-file ZIP archive hosted on this Mendeley Data record. Each expensive step checkpoints intermediate artefacts (odd-pair distance matrices as .npy files), so the pipeline is fully resumable after interruption.
Institutions
- Texas Tech UniversityTexas, Lubbock