Credit Card Fraud Detection

This is a machine learning project I built to detect fraudulent credit card transactions in real-time. It uses a Random Forest model hooked up to a Flask API, and includes a web interface where you can test out transactions to see if the model flags them as fraud.

How it works

The system is pretty straightforward:

• I trained a few machine learning models (using GridSearchCV for tuning) and saved the best one.
• There's a local MySQL database setup to store the prediction logs for analytics.
• The backend is a Flask app (app.py) that handles incoming transaction data and runs it through the model.
• The frontend (templates/index.html) is a simple UI for submitting transaction data manually.

Project Structure

• app.py: Main Flask application and prediction logic
• sql/: SQL scripts to create the database tables
• models/: Where the trained .pkl models are stored
• notebooks/: Jupyter notebooks I used for EDA and model training
• templates/: HTML files for the web interface
• render.yaml: Configuration for easy deployment on Render

Setup

1. Clone the repo

git clone https://github.com/avyasaini/Credit-Card-Fraud-Detection.git
cd Credit-Card-Fraud-Detection

2. Install dependencies

pip install -r requirements.txt

3. Run the app locally

python app.py

The app will run on http://127.0.0.1:5000.

API Usage

You can also test the /predict endpoint directly using Postman or curl. Send a POST request with JSON data like this:

{
  "amt": 281.06,
  "city_pop": 885,
  "lat": 35.9946,
  "long": -81.7266,
  "merch_lat": 36.430124,
  "merch_long": -81.179483,
  "unix_time": 1325466397,
  "category": "grocery_pos"
}

Performance

The Random Forest model performed really well on the validation set, achieving an F1-Score of around 0.99. I used balanced class weights to deal with the heavily imbalanced dataset (since most transactions aren't fraud).

Contact

If you have any questions or feedback about this project, feel free to open an issue in the repository!