In recent years, the rapid development of blockchain technologies has attracted considerable attention. However, little effort has been devoted toward investigating the large amount of trade data recorded in blockchains. This paper focuses on transaction data in Ethereum, which is a prominent public blockchain platform supporting not only secure cryptocurrency transfer but also various decentralized applications. By means of the framework of network science theory, we find that several transaction features, such as transaction volume, transaction relation, and component structure, exhibit a heavy-tailed property and can be approximated by the power law function. In particular, we find that the transaction relations follow a bow-tie structure with negative assortativity if they are regarded as a directed graph. The popular hubs tend to connect to a large number of common users. We believe that the aforementioned statistics can be ascribed to the vast diversity of transactions and the existence of a number of cryptocurrency exchanges. To the best of our knowledge, this study is the first to not only carry out a relatively comprehensive investigation of the transaction data recorded in Ethereum but also probe the statistical laws underlying the transaction relationships from the perspective of network science.