Bitcoin Address Behavior Dataset for Pattern Analysis

This study explores the application of machine learning algorithms to analyze behavioral patterns in Bitcoin addresses. Utilizing a dataset that includes various transaction details, we employ K-Nearest Neighbors (KNN), XGBoost, Random Forest (RF), and K-best feature selection as our base algorithms for initial pattern recognition. To enhance the predictive accuracy, we extend our analysis with advanced techniques including Stacking Classifier, Convolutional Neural Network (CNN), and Gradient Boosting Classifier. The primary objective is to accurately categorize Bitcoin addresses into specific categories such as blackmail, cyber-security service, darknet market, centralized exchange, P2P financial infrastructure service, gambling, money laundering, and more. This research aims to provide a robust framework for identifying suspicious activities and improving the security and transparency of Bitcoin transactions. The outcomes could significantly aid in the detection of illicit activities and the enforcement of cybersecurity measures in the cryptocurrency domain.

Keywords: Stacking Classifier, Convolutional Neural Network (CNN), and Gradient Boosting Classifier, classification algorithms, Kaggle dataset.