Scholarlink

<p>Bitcoin, introduced in 2008 by Satoshi Nakamoto, revolutionized the digital economy by enabling decentralized value storage and transfer, eliminating the need for a central authority. This paper presents a large-scale, temporally annotated graph dataset representing Bitcoin transactions, designed to advance research in blockchain analytics and beyond. The dataset comprises <b>252 million nodes</b> and <b>785 million edges</b>, with each node and edge timestamped for temporal analysis. To support supervised learning, we provide two labeled subsets: (i) <b>34,000 nodes</b> annotated with entity types, and (ii) <b>100,000 Bitcoin addresses</b> labeled with entity names and types. This dataset is the largest publicly available resource of its kind, addressing the limitations of existing datasets and enabling advanced exploration of Bitcoin’s transaction network. We establish baseline performance using graph neural network models for node classification tasks. Furthermore, we highlight additional use cases, including fraud detection, network analysis, and temporal graph learning, demonstrating its broader applicability beyond Bitcoin. We release the complete dataset, along with source code and benchmarks, to the public.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Bitcoin research with a transaction graph dataset

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年3月8日

查看全文
  • Hugo Schnoering,
  • Michalis Vazirgiannis

摘要

Bitcoin, introduced in 2008 by Satoshi Nakamoto, revolutionized the digital economy by enabling decentralized value storage and transfer, eliminating the need for a central authority. This paper presents a large-scale, temporally annotated graph dataset representing Bitcoin transactions, designed to advance research in blockchain analytics and beyond. The dataset comprises 252 million nodes and 785 million edges, with each node and edge timestamped for temporal analysis. To support supervised learning, we provide two labeled subsets: (i) 34,000 nodes annotated with entity types, and (ii) 100,000 Bitcoin addresses labeled with entity names and types. This dataset is the largest publicly available resource of its kind, addressing the limitations of existing datasets and enabling advanced exploration of Bitcoin’s transaction network. We establish baseline performance using graph neural network models for node classification tasks. Furthermore, we highlight additional use cases, including fraud detection, network analysis, and temporal graph learning, demonstrating its broader applicability beyond Bitcoin. We release the complete dataset, along with source code and benchmarks, to the public.