Quantum Machine Learning for Credit Card Fraud Detection

final product video walkthrough: https://www.youtube.com/watch?v=TSL_RbNKyI8&ab_channel=ryanWang

Overview

This project compares classical models (Logistic Regression, SVM) with quantum approaches, particularly the Variational Quantum Classifier (VQC), for binary fraud detection. We aim to evaluate whether quantum models can match or exceed classical performance on realistic financial data.

Goals

• Train classical and quantum classifiers on real-world data.
• Evaluate using recall, F1-score, and AUPRC.
• Analyze quantum-specific constraints like circuit depth and qubit limits.

Dataset

We use the Credit Card Fraud Detection dataset from Kaggle, containing:

• 284,807 transactions (492 fraud)
• Features: Time, Amount, V1–V28 (PCA), and Class (1 = fraud)

📥 Download Instructions

1. Sign in at Kaggle and go to the dataset page.
2. Click Download, unzip it, and place creditcard.csv in your project folder.

Alternatively (via Kaggle API):

pip install kaggle
kaggle datasets download -d mlg-ulb/creditcardfraud
unzip creditcardfraud.zip