This report takes the customer information and transaction information of 99bill.com as the research sample，considering that the existing system has a large amount of suspicious transaction warnings and a large amount of manual screening work，this paper uses Logistic regression scoring card model，combined with customer’s basic information and transaction data，analyzes the reported suspicious transactions，explores the commonality of suspicious transaction behavior，reconstructs suspicious transaction model monitoring model，and improves the monitoring and early warning processing effectiveness of third-party payment institutions suspicious transactions. On this basis，the paper puts forward some countermeasures and suggestions to further improve the anti-money laundering suspicious transaction monitoring and analysis system by using machine learning.