https://github.com/krishaa1803/bitcoin-transaction-anomaly-detection-using-unsupervised-machine-learning

Built an unsupervised Machine Learning pipeline to detect anomalies in Bitcoin transactions by selecting 19 key features from 700.
https://github.com/krishaa1803/bitcoin-transaction-anomaly-detection-using-unsupervised-machine-learning

anomaly-detection bitcoin blockchain data-science dbscan fraud-detection isolation-forest k-means machine-learning pca python tsne unsupervised-learning

Last synced: 2 months ago
JSON representation

Built an unsupervised Machine Learning pipeline to detect anomalies in Bitcoin transactions by selecting 19 key features from 700.

• Host: GitHub
• URL: https://github.com/krishaa1803/bitcoin-transaction-anomaly-detection-using-unsupervised-machine-learning
• Owner: krishaa1803
• License: mit
• Created: 2025-06-22T19:20:03.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2025-06-22T19:34:43.000Z (about 1 year ago)
• Last Synced: 2025-06-22T20:30:03.656Z (about 1 year ago)
• Topics: anomaly-detection, bitcoin, blockchain, data-science, dbscan, fraud-detection, isolation-forest, k-means, machine-learning, pca, python, tsne, unsupervised-learning
• Language: HTML
• Homepage:
• Size: 16.6 MB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Bitcoin-Transaction-Anomaly-Detection-using-Unsupervised-Machine-Learning

Built an unsupervised Machine Learning pipeline to detect anomalies in Bitcoin transactions by selecting 19 key features from 700. Used PCA, t-SNE for dimensionality reduction, Isolation Forest for anomaly detection, and K-Means/DBSCAN for clustering. Applied Hampel filter for noise correction and evaluated performance using Random Forest-derived silhouette scores.

---

## 🧠 Key Concepts

- **Unsupervised Learning**: No labeled data required.

- **Dimensionality Reduction**: Visualization and structure discovery.

- **Clustering & Isolation**: Identify anomalous transactions.

- **Feature Analysis**: Understand key drivers of anomalies.

---

## 🚀 Technologies & Libraries

- Python 3.x

- NumPy / Pandas

- Scikit-learn

- Matplotlib / Seaborn

- t-SNE / PCA

- Isolation Forest / DBSCAN / K-Means

- Hampel Filter for outlier preprocessing

---

## 📊 Pipeline Overview

### 1. 📂 Data Preprocessing

- Transaction data is cleaned and normalized.

- **Hampel filter** is applied to remove extreme outliers and reduce noise.

### 2. 🔻 Dimensionality Reduction

- **PCA** is used to reduce feature space while retaining variance.

- **t-SNE** helps in visualizing complex, high-dimensional patterns.

### 3. 📌 Clustering for Pattern Discovery

- **K-Means Clustering** for identifying common behavior groups.

- **DBSCAN** for density-based anomaly detection and noise separation.

- **Silhouette Score** is used to evaluate cluster quality.

### 4. 🚨 Outlier Detection

- **Isolation Forest** detects anomalous transactions by isolating rare patterns.

### 5. 📈 Feature Importance

- A **Random Forest** model ranks the most influential features post-clustering to help interpret anomaly causes (e.g., transaction value, frequency, mining difficulty, sentiment metrics).

---