Contextual Graph ETA Engine (CGEE™)

Overview

The Contextual Graph ETA Engine (CGEE™) is a graph-based, multi-horizon traffic ETA prediction system that integrates machine learning and shortest-path routing to generate realistic, context-aware travel time estimates.

The system combines:

• OpenStreetMap graph routing (OSMnx)
• Multi-horizon XGBoost speed prediction models
• Segment-level ETA aggregation
• Context-aware incident modeling
• Uncertainty band estimation
• Confidence scoring
• Route caching optimization

---

Key Features

• Graph-based shortest path routing
• Segment-level ML speed prediction
• Multi-horizon ETA forecasting (1h, 2h, 4h)
• Localized incident simulation with spatial propagation
• ETA uncertainty bands
• Confidence score per horizon
• Production-ready FastAPI backend
• Modular and extensible architecture

---

Architecture

User → FastAPI → Graph Routing → ML Speed Prediction → ETA Aggregation → Response (JSON)

---

API Endpoint

POST /predict-route-eta

Request

{
  "source": {"lat": 12.9716, "lon": 77.5946},
  "destination": {"lat": 12.9750, "lon": 77.6000}
}

{
  "eta_minutes": {
    "1_hour": {
      "estimate": 18.2,
      "lower_bound": 16.9,
      "upper_bound": 20.1
    }
  },
  "confidence": {
    "1_hour": 0.89
  },
  "meta": {
    "distance_km": 12.4,
    "segments": 48,
    "incident": true,
    "incident_segments": 4,
    "avg_incident_severity": 0.47
  }
}