This research applies machine learning for node classification in blockchain, revealing Neural Network's superior accuracy and fraud detection capabilities.
Key Points
Neural Network demonstrated a high accuracy of 95.3% for classifying blockchain nodes, significantly outperforming Random Forest and XGBoost.
Feature importance analysis identified Block Score and Transaction Fee as the most influential factors in predicting node behavior.
Machine learning models were evaluated on a dataset of 10,000 transactions with 16 attributes, showcasing the need for scalable fraud detection techniques.
Challenges remain in real-world deployment, highlighting the importance of real-time detection and adaptability against evolving fraud strategies.