BelBusRadar

Map View
Full Screen Map View
Smart Journey Planner
Journey Updates

Inspiration

As Belfast commuters ourselves, we were frustrated by needing 3+ separate apps just to plan a bus journey, check if it's on time, and buy a ticket. None of them show you safety information — crime hotspots, crash history, broken streetlights — things that actually matter when you're walking to a bus stop at night. We wanted to build the app we wished existed.

What it does

BelBusRadar is a real-time interactive map dashboard that combines live bus tracking, journey planning, route exploration, service updates, and AI-powered safety intelligence into one unified platform. It overlays 6 live data sources onto an interactive Mapbox map — bus stops, live moving buses, traffic crashes, potholes, street crime, and air quality. The "Make My Journey Safer" feature analyses real PSNI crime data within 100m of your route using weighted severity scoring to generate contextual safety tips.

How we built it

Frontend: React 19 + Vite + Tailwind CSS with Mapbox GL JS for the interactive map
Backend: Express.js server acting as an API proxy for the Translink EFA API (protecting the API key server-side) and the PSNI Crime API
Data integration: 6 live data sources — Translink departures/routes/journey planning, UK Police street crime, PSNI collision data, pothole reports, air quality stations, and lighting faults
Safety engine: Haversine distance filtering (100m radius), transit-relevant crime filtering, weighted severity scoring (robbery 4x vs anti-social behaviour 0.3x), time-of-day context
Design: Figma AI for initial UI wireframes, exported and integrated into the codebase. Custom canvas-drawn map icons since Mapbox doesn't support emoji in symbol layers
AI-assisted development: Claude Opus for agent-driven development, spec-driven methodology

Challenges we faced

Translink API documentation was a 185-page PDF — we had to extract it programmatically and reverse-engineer the EFA endpoint structure
Mapbox GL JS doesn't render emoji in symbol layers — we pivoted to canvas-drawn custom icons (triangles, X marks, diamonds)
The LineList endpoint returns 2MB of XML for 2,600+ routes — we added server-side caching and client-side pagination to keep it performant
Crime data volume — the PSNI Crime API returns 1,400+ crimes per month for Belfast city centre. We had to implement haversine distance filtering and weighted scoring to make the safety analysis meaningful rather than just scary
No live vehicle GPS — Translink's API is a journey planner, not a vehicle tracker. We interpolated bus positions between stops based on departure times using maths

What we learned

How to work with the EFA (Electronic Fahrplan Auskunft) journey planner API standard used across European transit systems
That combining multiple open data sources creates far more value than any single source alone
The importance of contextualising data — raw crime counts are meaningless without radius filtering, category weighting, and per-day normalisation
Canvas API for generating custom Mapbox marker images at runtime