This project aims to develop a comprehensive analytics system to enhance Anti-Money Laundering (AML) compliance on cryptocurrency transaction platforms. By leveraging data engineering techniques, machine learning algorithms, and interactive data visualizations, the system efficiently detects suspicious transaction patterns and identifies high-risk users.

• 203,768 Transactions analyzed
• 1,000 Users profiled
• 14 Suspicious users identified
• 42,019 Suspicious transactions detected

Analysis indicates that 77.1% of transactions are unclassified (unknown), 20.6% are identified as illicit, and only 2.23% are labeled as licit. The high proportion of unclassified and illicit transactions underscores the necessity of sophisticated detection systems.

The analysis reveals significant fluctuations in transaction volume over time, with peaks reaching approximately 8,000 transactions and troughs around 1,000 transactions. Identifying unusual spikes aids in detecting potential coordinated suspicious activities.

Transaction distribution per user displays a bimodal pattern, where most users conduct very few transactions, while a small group of users execute a significantly higher number of transactions (800+ transactions). This pattern suggests possible structuring or smurfing activities.

Our anomaly detection system identifies 1.4% of users as potentially suspicious by utilizing multiple machine learning algorithms, including:

• Isolation Forest
• Local Outlier Factor (LOF)
• One-Class SVM

The system successfully flagged 14 suspicious users with anomalous transaction patterns. These users executed 816 transactions within a single day, indicating abnormally high transaction frequency.

This project employs a multi-faceted approach to AML compliance:

• Data Integration – Merging transaction data with user KYC information
• Feature Engineering – Creating behavioral indicators for anomaly detection
• Machine Learning – Implementing multiple anomaly detection algorithms
• Ensemble Techniques – Combining model outputs for higher accuracy
• Risk Scoring – Developing a structured risk assessment framework
• Interactive Visualization – Building a real-time monitoring dashboard

A Streamlit dashboard has been developed to assist compliance officers in:

• Filtering transactions by time period, labels, and KYC levels
• Visualizing transaction patterns and anomalies
• Investigating specific suspicious users
• Downloading reports for further analysis

For in-depth analysis, refer to the following reports:

• Anomaly Detection Report – Insights into anomaly detection methodologies and results
• Risk Assessment Report – Comprehensive breakdown of risk scoring methods
• User Profiling Report – Identification of behavioral patterns and suspicious activities
• Overview Report – Executive summary of key findings

• Python – Core programming language
• Pandas & NumPy – Data manipulation and analysis
• Scikit-learn – Machine learning algorithms
• dbt – Data transformation workflows
• Plotly & Matplotlib – Data visualization
• Streamlit – Interactive dashboard development
• SQL – Database querying and modeling

• Real-time Processing – Implementing stream processing for live transaction monitoring
• Network Analysis – Graph-based analysis of transaction networks
• Explainable AI – Enhancing interpretability of anomaly detection models
• Regulatory Reporting – Automating compliance reporting for authorities
• Alert Management – Developing a case management system for alert investigations

For more information, please reach out to:

Name: Nugrah Salam
Email: ompekp@gmail.com
GitHub: envexx

This project was developed as a portfolio piece showcasing advanced data analytics capabilities in cryptocurrency compliance.