Blockchain abnormal transaction detection method based on weighted sampling neighborhood nodes

Abstract

In view of the increasingly serious problem of money laundering crime in the blockchain industry, the existing solutions in this scenario can not be applied to reality, or there is a high false positive rate and false negative rate, a method based on weighted sampling neighborhood nodes is designed to find and analyze the implied interrelationship in the data between blockchain transaction features, and learn more effective aggregate input features in the local neighborhood of nodes through model training. Good results have been achieved on the public data set, which verifies the effectiveness of the model in the field of blockchain abnormal transaction detection. Moreover, it provides an idea of modelling transaction entity data by using graph neural network structure for the field of financial data transactions such as anti-money laundering monitoring.