Clustering Blockchain Data

Abstract

Blockchain datasets, such as those generated by popular cryptocurrencies Bitcoin, Ethereum, and others, are intriguing examples of big data. Analysis of these datasets has diverse applications, such as detecting fraud and illegal transactions, characterizing major services, identifying financial hotspots, and characterizing usage and performance characteristics of large peer-to-peer consensus-based systems. Unsupervised learning methods in general, and clustering methods in particular, hold the potential to discover unanticipated patterns leading to valuable insights. However, the volume, velocity, and variety of blockchain data, as well as the difficulties in evaluating results, pose significant challenges to the efficient and effective application of clustering methods to blockchain data. Nevertheless, recent and ongoing work has adapted classic methods, as well as developed new methods tailored to the characteristics of such data. This chapter motivates the study of clustering methods for blockchain data, and introduces the key blockchain concepts from a data-centric perspective. It presents different models and methods used for clustering blockchain data, and describes the challenges and some solutions to the problem of evaluating such methods.