Identifying User Behavior Profiles in Ethereum Using Machine Learning Techniques

Abstract

Ethereum is one of the largest blockchain platforms currently that has become a digital business environment for users. This platform is designed to allow decentralized transactions between anonymous users. Thus, the development of methods to identify user behavior profiles, keeping their identities anonymous, has the potential to leverage business on this platform. In this work, we investigate the use of machine learning to classify a user profile as professional or common based on the attributes of their transactions. This classification is challenging due to the small fraction of publicly labeled users in Ethereum and still the considerably smaller fraction of professional users. To conduct this investigation, we train models considering carefully balanced sets of transactions with labeled users. Our results show high performance models for the classification of profiles, achieving a performance greater than 90% for accuracy, precision, and other related measures. In addition, we have identified the most relevant features in transactions for this classification.