London Synergy Index
Predictive Site Selection
The London Synergy Index is a geospatial intelligence platform that identifies underserved retail locations across Greater London. By tessellating the city into 55,000 H3 hexagons and enriching each with satellite-derived population data, census demographics, and POI density features, the system builds a comprehensive spatial understanding of commercial opportunity. A spatially cross-validated XGBoost classifier then scores each hexagon for investment potential.
Key Features
Spatial Feature Engineering
Each H3 hexagon is enriched with multi-source features: LandScan satellite rasters provide population estimates, ONS Census data adds demographic dimensions (income, age, employment), and OpenStreetMap POIs are processed through a NetworkX graph to compute centrality and competition-density metrics. These features are combined into a unified feature matrix for modeling.
Modeling Approach
The classification pipeline uses XGBoost with spatial cross-validation — a critical design decision that prevents spatial autocorrelation from inflating model performance. Traditional k-fold CV would leak spatial information between folds; spatial CV ensures train/test splits respect geographic boundaries, producing realistic accuracy estimates.
Structural Hole Analysis
Burt’s Structural Hole Theory from network science is adapted to identify gaps in London’s retail network. By modeling retail locations as nodes in a spatial graph, the system detects areas where new businesses could bridge disconnected clusters — these structural holes represent high-potential investment opportunities.