💳 Credit Card Fraud Detection using DBSCAN Clustering

An unsupervised machine learning approach to detect fraudulent credit card transactions using DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm — no labeled training data required.

📋 Table of Contents

• Overview
• Demo
• Features
• Installation
• Usage
• Dataset
• Methodology
• Results
• Contributing
• License
• Contact

🎯 Overview

This project implements an unsupervised anomaly detection system for identifying potentially fraudulent credit card transactions. Unlike supervised approaches, DBSCAN doesn't require labeled training data and can identify outliers (potential fraud) based on transaction patterns and density.

Why DBSCAN?

• No labeled data required - Works as an unsupervised approach
• Identifies outliers - Transactions marked as noise (-1) may indicate fraud
• Handles varying density - Can detect fraud patterns of different shapes
• Robust to noise - Doesn't force every point into a cluster

🖼️ Demo

✨ Features

• 📊 Automated Data Processing - Handles missing values and infinite entries
• 🔄 Feature Standardization - Normalizes data for optimal clustering
• 🎯 DBSCAN Clustering - Density-based anomaly detection
• 📈 PCA Visualization - 2D projection of high-dimensional data
• 🧩 Silhouette Score - Quality metric for cluster separation
• 🎨 Plots - Visual representation of clusters and outliers

🚀 Installation

Prerequisites

• Python 3.7 or higher
• Google Colab (recommended) or local Jupyter environment

Dependencies

pip install pandas numpy matplotlib scikit-learn

Clone Repository

git clone https://github.com/zain-cs/Credit-Card-Fraud-DBSCAN.git cd Credit-Card-Fraud-DBSCAN

💻 Usage

Running in Google Colab

1. Upload DBSCAN_clustering.ipynb to Google Colab
2. Run all cells sequentially
3. Upload your credit card transaction CSV when prompted
4. View results and visualizations

Running Locally

1. Open DBSCAN_clustering.ipynb in Jupyter Notebook/Lab
2. Place your dataset CSV in the same directory (or update the file path in the notebook)
3. Run all cells

Repository Structure

Credit-Card-Fraud-DBSCAN/ │ ├── DBSCAN_clustering.ipynb # Main notebook: preprocessing, clustering, evaluation ├── README.md # This file └── LICENSE # MIT License

📊 Dataset

This project works with credit card transaction datasets containing numerical features. The expected format:

• Features: V1, V2, ..., V28 (PCA-transformed features)
• Time: Seconds elapsed between transactions
• Amount: Transaction amount
• Class (optional): 0 = Normal, 1 = Fraud (for validation only)

Recommended Dataset

Kaggle Credit Card Fraud Detection Dataset

Data Format

Time,V1,V2,V3,...,V28,Amount,Class 0,-1.359807134,-0.072781,...,149.62,0 1,-1.358354062,1.191857,...,2.69,0

🔬 Methodology

1. Data Preprocessing

• Remove infinite and missing values
• Select only numerical features
• Standardize features (mean=0, std=1)

2. DBSCAN Clustering

DBSCAN(eps=1.5, min_samples=5)

• eps: Maximum distance between neighbors (tunable)
• min_samples: Minimum points to form a dense region

3. Anomaly Detection

• Points labeled as -1 are outliers (potential fraud)
• Dense clusters represent normal transaction patterns
• Noise points deviate significantly from normal behavior

4. Evaluation

• Silhouette Score: Measures cluster cohesion (excluding noise)
• Cluster Distribution: Shows normal vs. outlier counts
• PCA Visualization: 2D representation of clusters

📈 Results

📊 Cluster Distribution: 0 280000 -1 1500 1 500

• Cluster 0: Normal transactions (majority)
• Cluster -1: Outliers (potential fraud)
• Other clusters: Distinct transaction patterns

Interpretation

🔴 High Priority: Transactions in cluster -1 (noise/outliers)
🟡 Medium Priority: Small, isolated clusters
🟢 Low Priority: Large, dense clusters

⚙️ Hyperparameter Tuning

Adjust DBSCAN parameters for your dataset:

More sensitive to outliers

DBSCAN(eps=0.5, min_samples=3)

More conservative

DBSCAN(eps=2.0, min_samples=10)

Guidelines:

• Smaller eps → More outliers detected (higher sensitivity)
• Larger min_samples → Fewer small clusters (more conservative)

🤝 Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

📧 Contact

Zain
GitHub: @zain-cs

🙏 Acknowledgments

• Scikit-learn for machine learning tools
• Kaggle for providing datasets

📚 References

• Ester, M., et al. (1996). "A density-based algorithm for discovering clusters"
• Credit Card Fraud Detection Dataset (Kaggle)
• Scikit-learn DBSCAN Documentation

---

⭐ If you found this project helpful, please consider giving it a star!