README.md

metadata

license: mit
title: Fraud_Detection_using_ML
sdk: docker
emoji: 🏢
colorFrom: green
colorTo: red
pinned: true

Fraud Detection System

An interactive web application for real-time fraud detection using machine learning.

Features

• Real-time Analysis: Instant fraud detection with probability scores
• Interactive Interface: User-friendly web interface with Bootstrap styling
• Machine Learning: Random Forest classifier trained on transaction data
• Risk Assessment: 5-level risk classification (Very Low to Very High)
• Visual Results: Progress bars and color-coded risk indicators

Installation

1. Install Python dependencies:

   pip install -r requirements.txt
   

2. Run the application:

   python app.py
   

3. Or use the batch file (Windows):

   start_app.bat
   

Usage

1. Open your browser and go to http://localhost:5000
2. Fill in the transaction details:
   • Personal Information (Gender, Age, Location)
   • Account Information (Bank Branch, Account Type, Balance)
   • Transaction Information (Amount, Type, Device, etc.)
3. Click "Analyze Transaction"
4. View the fraud probability and risk assessment

Input Fields

Personal Information

• Gender: Female or Male
• Age: Age of the account holder (18-100)
• State: Indian states (Andhra Pradesh to Ladakh)
• City: Major Indian cities (Mumbai, Delhi, Bangalore, etc.)

Account Information

• Bank Branch: Indian bank branches (SBI, HDFC, ICICI, etc.)
• Account Type: Savings, Current, Credit Card, Fixed Deposit, Salary, Joint
• Account Balance: Current account balance in Indian Rupees (₹)

Transaction Information

• Transaction Amount: Amount being transacted in Indian Rupees (₹)
• Transaction Type: ATM Withdrawal, Bank Deposit, Fund Transfer, UPI Payment, etc.
• Transaction Device: ATM, POS, Mobile Apps, UPI platforms, Internet Banking, etc.
• Transaction Date: Day of month (1-31)
• Transaction Time: Hourly time slots (24-hour format)
• Currency: INR (default), USD, EUR, GBP, AED, SGD, CAD, AUD

API Endpoints

GET /

Returns the main web interface

POST /predict

Request Body:

{
    "gender": "1",
    "age": "35",
    "state": "15",
    "city": "127",
    "bank_branch": "127",
    "account_type": "2",
    "transaction_date": "22",
    "transaction_time": "52151",
    "transaction_amount": "32415.45",
    "transaction_type": "3",
    "account_balance": "74557.27",
    "transaction_device": "17",
    "transaction_currency": "0"
}

Response:

{
    "prediction": 0,
    "fraud_probability": 15.75,
    "risk_level": "Very Low",
    "risk_color": "#28a745",
    "message": "Legitimate Transaction"
}

Risk Levels

• Very Low (0-20%): Green (#28a745)
• Low (20-40%): Teal (#20c997)
• Medium (40-60%): Yellow (#ffc107)
• High (60-80%): Orange (#fd7e14)
• Very High (80-100%): Red (#dc3545)

Testing

Run the API test script:

python test_api.py

Files Structure

fraud_flask_app/
├── app.py                 # Main Flask application
├── model.pkl             # Trained Random Forest model
├── encoders.pkl          # Label encoders for categorical features
├── requirements.txt      # Python dependencies
├── start_app.bat        # Windows startup script
├── test_api.py          # API testing script
├── templates/
│   └── index.html       # Web interface template
├── static/
│   └── style.css        # Custom styling
└── README.md           # This file

Model Information

• Algorithm: Random Forest Classifier
• Features: 13 input features
• Training Data: Bank transaction fraud dataset
• Performance: Optimized for fraud detection accuracy

Technologies Used

• Backend: Flask (Python)
• Frontend: HTML5, CSS3, JavaScript, Bootstrap 5
• Machine Learning: scikit-learn, pandas, numpy
• Icons: Font Awesome

Browser Support

• Chrome/Chromium (recommended)
• Firefox
• Safari
• Edge

Troubleshooting

Model Loading Issues

• Ensure model.pkl and encoders.pkl are in the same directory as app.py
• Check Python package versions match requirements.txt

Connection Issues

• Verify Flask is running on port 5000
• Check firewall settings
• Ensure no other application is using port 5000

Prediction Errors

• Validate all input fields are filled correctly
• Check input value ranges match the specified limits
• Ensure numeric fields contain valid numbers

License

This project is for educational and demonstration purposes.