Production Setup - Fraud Detection System

Overview

This guide covers the production deployment setup for the fraud detection ML system. The application is designed to run with Gunicorn as the WSGI HTTP server instead of Flask’s development server.

Production Server Configuration

Using Gunicorn

The application uses Gunicorn (Green Unicorn) as the production WSGI HTTP server. This is defined in the Procfile:

web: gunicorn main:app

This configuration:

Runs Gunicorn as the web server
Points to the app object in main.py
Uses default Gunicorn workers and settings

Why Gunicorn?

Flask’s built-in development server is not suitable for production because:

It’s single-threaded and cannot handle concurrent requests efficiently
It lacks production-grade security features
It’s not optimized for performance

Gunicorn provides:

Multiple worker processes for handling concurrent requests
Production-ready performance and stability
Better resource management

Environment Variables

PORT Configuration

The application reads the PORT from environment variables with a fallback default:

port = int(os.getenv("PORT", 5001))

Environment Variables:

PORT: The port number for the application (default: 5001)
LANG: Set to en_US.UTF-8 for proper encoding
LC_ALL: Set to en_US.UTF-8 for locale settings

Setting environment variables:

export PORT=8000
export LANG=en_US.UTF-8
export LC_ALL=en_US.UTF-8

File Structure Requirements

The application expects the following directory structure:

project-root/
├── main.py                          # Main Flask application
├── Procfile                         # Process configuration
├── requirements.txt                 # Python dependencies
├── templates/                       # HTML templates
│   └── index.html
├── Training_Batch_Files/            # Training data batches
├── Prediction_Batch_files/          # Prediction data batches
├── Training_Raw_data_validation/    # Training data validation
├── Prediction_Raw_Data_Validation/  # Prediction data validation
├── data/                            # Data storage
├── application_logging/             # Application logs
└── flask_monitoringdashboard.db     # Monitoring database

Required Modules

Ensure these custom modules are present:

prediction_Validation_Insertion.py
trainingModel.py
training_Validation_Insertion.py
predictFromModel.py
DataTransform_Training/
DataTransformation_Prediction/
DataTypeValidation_Insertion_Training/
DataTypeValidation_Insertion_Prediction/
data_ingestion/
data_preprocessing/
file_operations/
best_model_finder/

Database Setup

Flask Monitoring Dashboard Database

The application uses SQLite for the Flask Monitoring Dashboard:

Database file: flask_monitoringdashboard.db
Automatically created on first run
Stores monitoring metrics and performance data

Schema Files

The system uses JSON schema files for data validation:

schema_training.json: Training data schema
schema_prediction.json: Prediction data schema

Ensure these files are present and properly configured before deployment.

Dependencies Installation

Install all required dependencies from requirements.txt:

pip install -r requirements.txt

Key dependencies:

Flask 1.1.1
Flask-Cors 3.0.8
Flask-MonitoringDashboard 3.0.6
Gunicorn 20.0.4
scikit-learn 0.22.1
pandas 0.25.3
numpy 1.18.1
xgboost 0.90
imbalanced-learn 0.6.1

Running in Production

Local Production Testing

Test the production configuration locally:

gunicorn main:app --bind 0.0.0.0:8000

With Worker Processes

For better performance, specify multiple workers:

gunicorn main:app --bind 0.0.0.0:8000 --workers 4

With Logging

gunicorn main:app \
  --bind 0.0.0.0:8000 \
  --workers 4 \
  --access-logfile - \
  --error-logfile -

Security Considerations

CORS Configuration: The application uses Flask-CORS with open access. Review and restrict for production:
```
CORS(app, resources={r"/*": {"origins": "https://yourdomain.com"}})
```
Debug Mode: Ensure debug=True is removed or set to False in production
Environment Variables: Use environment-specific configuration files
File Permissions: Restrict access to data directories and log files

Health Check

The home route serves as a basic health check:

curl http://localhost:8000/

Should return the index.html template.

Next Steps

Configure monitoring dashboard
Deploy to Heroku
Set up automated backups for training data
Configure logging aggregation

Production

​Overview

​Production Server Configuration

​Using Gunicorn

​Why Gunicorn?

​Environment Variables

​PORT Configuration

​File Structure Requirements

​Required Modules

​Database Setup

​Flask Monitoring Dashboard Database

​Schema Files

​Dependencies Installation

​Running in Production

​Local Production Testing

​With Worker Processes

​With Logging

​Security Considerations

​Health Check

​Next Steps

Build docs developers (and LLMs) love