SmartGov AI

This project is a fraud detection system for government schemes. It uses a RandomForestClassifier to detect potential fraud based on applicant data and serves predictions via a FastAPI backend.

Structure

• data/: Contains the synthetic dataset (fraud_dataset.csv).
• models/: Stores the trained machine learning model (fraud_model.pkl).
• backend/: Contains the FastAPI application (main.py).
• frontend/: Directory reserved for the frontend web application (currently empty as requested).
• utils/: Contains utility scripts:
  • generate_data.py: Script to generate the 300-row synthetic dataset.
  • train_model.py: Script to train the RandomForestClassifier on the generated data.

Getting Started

1. Install Dependencies

Make sure you have Python installed. Install the required packages:

pip install -r requirements.txt

2. Generate the Dataset

Run the data generation script to create a synthetic dataset of 300 scheme applicants:

python utils/generate_data.py

This will output data/fraud_dataset.csv.

3. Train the Model

Train the fraud detection model:

python utils/train_model.py

This will train the model and save it to models/fraud_model.pkl.

4. Configure Environment

To use the AI explanation feature, create a .env file from the example:

cp .env.example .env

Open .env and add your OpenAI API key.

5. Start the Backend API

Run the FastAPI application using Uvicorn:

uvicorn backend.main:app --reload

The API will be accessible at: http://127.0.0.1:8000.

6. Start the Frontend Dashboard (Optional)

Run the Streamlit UI to get a user-friendly dashboard:

streamlit run frontend/app.py

The dashboard will open automatically in your browser.

API Endpoints

• GET / — Health check endpoint.
• POST /predict — Used to check an applicant for fraud. Example payload:

  {
    "income": 25000,
    "family_size": 4,
    "location": "Delhi",
    "scheme_type": "PM_Kisan",
    "bank_account": "ACC999",
    "phone_number": "9998887776"
  }

• POST /explain — Submits a prediction result to generate a human-readable AI explanation. Example payload:

  {
    "income": 95000,
    "family_size": 1,
    "location": "Delhi",
    "scheme_type": "PM_Kisan",
    "fraud": 1,
    "confidence": 87.5,
    "anomaly_score": -0.12,
    "graph_cluster_size": 8,
    "graph_connected_fraud_rate": 75.0
  }