API Endpoints (LSTM Traffic Predictor)

Health & Information

GET /health

Service health check.

Response (200):

{
  "status": "healthy",
  "message": "LSTM service is ready"
}

GET /model-info

Retrieve LSTM model metadata.

Response (200):

{
  "model_type": "LSTM",
  "input_shape": [3, 5],
  "output_shape": [5],
  "description": "Predicts edge density for next hour based on 3 hourly measurements",
  "edges": ["-269002813", "-55825089", "617128762", "-617128762", "-312266114#2"],
  "test_loss": 0.0698,
  "test_mae": 0.2084
}

Traffic Density Prediction

POST /predict

Predict traffic density for the next hour.

Request:

{
  "data": [
    [18.93, 10.13, 5.23, 4.14, 3.08],
    [24.02, 11.01, 8.98, 5.42, 4.26],
    [22.14, 9.34, 5.62, 4.57, 3.81]
  ]
}

Parameters: - data (required): Array of 3 hourly measurements for 5 edges. Shape: (3, 5) - 3 timesteps (hours) - 5 edge density values per timestep

Response (200):

{
  "prediction": [25.47, 12.15, 9.23, 6.81, 5.14],
  "edge_ids": ["-269002813", "-55825089", "617128762", "-617128762", "-312266114#2"]
}

Error Responses:

400 Bad Request: Invalid shape or missing data field

{
  "detail": "Expected shape (3, 5), got (2, 5)"
}

503 Service Unavailable: Model not loaded
```
{
  "detail": "Model not loaded"
}
```

500 Internal Server Error: Prediction inference failed

{
  "detail": "Prediction error: [error details]"
}

Testing

Health Check

curl http://localhost:8001/health

Model Info

curl http://localhost:8001/model-info

Predict Density

curl -X POST \
  -H "Content-Type: application/json" \
  -d '{
    "data":[
      [18.93,10.13,5.23,4.14,3.08],
      [24.02,11.01,8.98,5.42,4.26],
      [22.14,9.34,5.62,4.57,3.81]
    ]
  }' \
  http://localhost:8001/predict

Notes

Port: 8001
Input Format: Raw (non-normalized) traffic density values
Output Format: Predicted density values for next hour (denormalized)
Normalization: Handled internally using MinMaxScaler
Top 5 Edges: Model trained on the 5 most congested edges from SUMO simulation