Credit Card Fraud Detection

A machine learning project implementing multiple approaches to detect fraudulent credit card transactions.

Project Overview

This project implements various machine learning models to detect credit card fraud using a dataset of transactions. The approaches include traditional ML models, anomaly detection, and deep learning techniques.

Dataset

The dataset contains credit card transactions made by European cardholders over two days in September 2013. Due to confidentiality issues, the original features have been transformed using PCA.

Dataset Characteristics:

• Features: V1-V28 (PCA components), Time, Amount, Class
• Binary Classification: 0 (Normal), 1 (Fraud)
• Highly Imbalanced Dataset
• All numerical features

Note: Due to the large size of the dataset (creditcard.csv), it is not included in the repository. You can download it from Kaggle Credit Card Fraud Detection Dataset.

Setup

1. Clone the repository

git clone https://github.com/shoryasethia/financial-fraud-detection.git
cd financial-fraud-detection

2. Download the dataset

• Download creditcard.csv from Kaggle
• Place it in the project root directory

3. Install dependencies

pip install -r requirements.txt

Models Implemented

1. Traditional Machine Learning

   • Random Forest Classifier
   • LightGBM
   • XGBoost

2. Anomaly Detection

   • Gaussian Anomaly Detection with Power Transforms
   • GridSearchCV optimization

3. Deep Learning

   • Undercomplete Autoencoder
   • Architecture: 128-64-32-16 with BatchNorm
   • Trained on non-fraudulent transactions

Results

• Autoencoder Performance:
  • F2 Score: 0.962
  • F1 Score: 0.917

Project Structure

financial-fraud-detection/
├── autoencoder.ipynb        # Autoencoder implementation
├── main.ipynb              # Traditional ML models
├── autoencoder.pth         # Saved model weights
├── requirements.txt        # Project dependencies
└── README.md              # Project documentation

Dependencies

• Python 3.x
• PyTorch
• scikit-learn
• pandas
• numpy
• seaborn
• matplotlib
• lightgbm
• xgboost