Safety Intelligence API
Predictive crime risk intelligence and spatial forecasting for global applications.
The Safety Intelligence API (SIA) delivers predictive crime risk intelligence through a sophisticated spatial forecasting system. Built from the ground up with South African safety challenges in mind, SIA is trained on data from cities across South Africa, America, and the UK, providing probability-based risk assessments for any covered location.
SIA is currently in internal beta for research and development. Access is restricted to approved partners and internal teams. Breaking changes may occur as we refine the platform.
Platform Capabilities
SIA is powered by a proprietary spatial convolutional network architecture that models crime patterns across space and time. The system learns from millions of historical records to forecast probability distributions for the week ahead.
The Community Wolf Dataset
SIA is trained on the Community Wolf dataset, a unique combination of open-source crime data and proprietary intelligence from our ecosystem. This includes data from our partner network and internal products such as Namola (South Africa's leading personal safety app), Community Wolf, and Community Wolf Business Tools, accounting for several million data points that are not available anywhere else. Our deep roots in the South African market mean we have unparalleled ground-truth data for local conditions, while our international coverage demonstrates the model's ability to generalise across different urban environments.
What SIA Delivers
| Capability | Description |
|---|---|
| Multi-City Coverage | 10 cities across 3 continents with 1,800+ spatial cells |
| Auto-Routing | Automatically detects which city covers any given coordinates |
| Probability Distributions | Full probability breakdown across 6 standardised crime categories |
| Temporal Forecasting | Predictions available for 40+ weeks into the future |
| Route Risk Analysis | Cumulative risk scoring along any path with segment-by-segment breakdown |
| Batch Processing | Query multiple locations across cities in a single request |
| Sub-10ms Latency | Optimised for real-time applications |
Standardised Risk Categories
SIA normalises crime data from every city into six consistent categories:
| Category | Description | Typical Distribution |
|---|---|---|
THEFT_BURGLARY | Theft, burglary, robbery | ~30% |
BATTERY_ASSAULT | Assault, battery, physical violence | ~27% |
VEHICLE_DAMAGE | Vehicle theft, vandalism, arson | ~17% |
OTHER | Narcotics, weapons, trespass | ~24% |
SERIOUS_VIOLENT | Homicide, sexual assault, kidnapping | ~1.4% |
NO_CRIME | No significant crime expected | Variable |
Model Performance
| Metric | Value | Interpretation |
|---|---|---|
| Accuracy | 71.1% | Correctly predicts dominant crime type 7 out of 10 times |
| MAE | 3.82% | Average probability prediction error |
| Brier Score | 0.0355 | Probability calibration quality (lower is better) |
| F1 Score | 0.697 | Balance of precision and recall |
Use Cases
SIA is designed for applications where location-based risk intelligence adds value:
- Navigation and routing: Integrate safety scoring into route planning and turn-by-turn guidance
- Personal safety applications: Alert users when entering elevated-risk areas
- Delivery and logistics: Route optimisation with safety scoring for drivers and couriers
- Real estate insights: Neighbourhood risk comparison for property decisions
- Insurance risk modelling: Location-based premium calculation
- Event security planning: Venue risk assessment and staffing decisions
- Fleet management: Driver safety scoring and parking risk analysis
Spatial Grid System
SIA uses H3 hexagonal indexing (Uber's spatial system) to divide cities into consistent ~5 km² cells. Each cell receives an independent probability forecast.
- Globally unique cell identifiers
- Consistent cell sizes worldwide
- Built-in hierarchy for zoom levels
- Efficient spatial queries
Endpoints Reference
Request and response documentation for all API endpoints
Authentication
Current access requirements and future plans
Beta Limitations
The current beta implementation has certain constraints:
- Partner access only: Not available for public use
- Static predictions: Refreshed weekly rather than real-time
- Coverage gaps: Not all areas within a city have full coverage
Roadmap
Future releases will expand SIA capabilities:
- Real-time prediction updates
- Time-of-day risk adjustments (day/night, weekday/weekend)
- Custom area polygon queries
- Historical trend analysis
- Webhook notifications for risk changes
- Additional city coverage
Interested in partnership or early access? Email us at hello@communitywolf.com