Research On Blockchain Transaction Behavior Prediction Method Based On Graph Feature Extraction

With the continuous development of digital currency,blockchain technology,as its underlying technology,is gradually applied to various industries,and a large number of blockchain transaction data will lead to the difficulty of monitoring the blockchain system.Blockchain transaction behavior prediction can effectively depict the transaction behavior of blockchain users,thus effectively detect abnormal transaction behavior in the blockchain,which has important theoretical and practical significance for monitoring malicious transactions in the blockchain and ensuring the security of the blockchain system.At present,scholars usually use random walk,cascade machine learning,graph neural network and other transaction graph-based methods to predict the transaction behavior of blockchain.However,the current problems such as the heterogeneity of blockchain data,the large volume of transaction graph and the difficulty in judging the importance of transaction characteristics limit the universality,efficiency and accuracy of blockchain transaction behavior prediction.To solve the above problems,this paper proposes a blockchain transaction behavior prediction method based on graph feature extraction.First,in view of the lack of universality caused by the heterogeneity of blockchain data,we propose a blockchain transaction graph construction method,which converts the heterogeneous blockchain transaction data into a common transaction graph structure;Then,in view of the problem that it is difficult to calculate due to the large volume of transaction graph and too many nodes,we propose a transaction subgraph division method based on node association and social network,which divides the huge transaction graph into several transaction subgraphs with strong correlation,thus improving the calculation efficiency;Finally,in view of the problem that the importance of transaction features is difficult to judge,we propose a blockchain transaction behavior prediction method based on graph feature extraction.By extracting the graph features of both parties in the transaction subgraph and filtering the features,we can predict the blockchain transaction behavior.By collecting real Bitcoin block data,we use the top 95% transaction data to predict the bottom 5% transactions in chronological order,with an accuracy of 81.12%.