Credit Card Fraud Detection Using Machine Learning

Overview

This project applies machine learning techniques to detect fraudulent credit card transactions using a publicly available dataset from Kaggle.

Final Report

Download the full report (PDF)

Dataset

• Source: Kaggle (ULB Machine Learning Group)
• Contains anonymized transaction data
• Highly imbalanced dataset with very few fraudulent transactions

Methods

• Exploratory Data Analysis (EDA)
• Feature selection and scaling
• Logistic Regression
• Random Forest

Evaluation

Models were evaluated using:

• Confusion Matrix
• Precision
• Recall
• F1-score

Results

• Logistic Regression performed well overall but missed some fraudulent transactions
• Random Forest achieved better performance, particularly in recall and F1-score
• Random Forest was more effective at detecting fraud

Key Takeaways

• Class imbalance significantly impacts model evaluation
• Accuracy alone is not a reliable metric
• Recall is critical for fraud detection problems

Repository Structure

• eda.ipynb – Data analysis and modeling
• capstone.png – Model comparison results
• LaTeX report – Final project report

Author

James D. Pinkston

Development Setup

For environment setup and development workflow instructions, see: SETUP.md