Blockchain Data Mining With Graph Learning: A Survey

Area of Science:

• Computer Science
• Data Science
• Cybersecurity

Background:

• Blockchain data mining offers insights into system operations and participant behavior.
• Traditional analysis methods struggle with the large volume and complex structure of blockchain data.
• Graph learning presents a powerful approach to analyze blockchain data by considering node features and relationships.

Purpose of the Study:

• To systematically review blockchain data mining tasks utilizing graph learning approaches.
• To provide a comprehensive overview of existing methods for blockchain data acquisition, graph construction, and feature extraction.
• To classify and compare graph learning algorithms applied to blockchain analysis.

Main Methods:

• Investigated blockchain data acquisition and sampling techniques (rule-based, cluster-based).
• Classified graph construction methods (transaction-based, account-based) and analyzed feature extraction.
• Compared graph learning algorithms (traditional ML, graph representation, graph deep learning) for blockchain applications.

Main Results:

• Identified key methods for blockchain data acquisition, graph construction, and feature extraction.
• Categorized graph learning algorithms applicable to blockchain data analysis.
• Highlighted the potential of graph learning to address challenges in blockchain data mining.

Conclusions:

• Graph learning is a promising approach for analyzing complex blockchain data.
• The review provides a structured overview of current graph-based techniques in blockchain data mining.
• Identified future research directions and open challenges in the field.