Event-Based ML Backend

This backend implements a data-driven ML system for automated income loss claims.

Prerequisites

• Python 3.8+
• Virtual environment (recommended)

Setup Instructions

1. Navigate to the backend directory:

   cd backend

2. Create and activate a virtual environment:

   python -m venv venv
   # On Windows:
   .\venv\Scripts\activate
   # On macOS/Linux:
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Initialize AI Models (CRITICAL): Run the training script to generate the regional models and initial synthetic data.

   python ai_service/train.py

   This will create a backend/ai_service/models/ folder with .pkl files.

Running the Application

1. Start the Flask Server:

   python app.py

   The server will start at http://localhost:5000.

2. Trigger the Weekly Pipeline (Admin Hook): You can manually trigger a full system evaluation (ingestion -> features -> AI -> decision -> claims) by sending a POST request to: http://localhost:5000/predictions/run-pipeline

Running Tests

To verify all API endpoints and logic:

python -m pytest tests/

Architecture

• ai_service/: Standalone ML training and prediction logic (Region-based & Personalized).
• services/: Core logic for ingestion, preprocessing, and the rules engine.
• routes/: Flask API endpoints.
• jobs/: Background tasks/pipelines.
• db/: Mock MongoDB implementation.