The Design And Implementation Of AML Rule Based Structural Test Data Generation System

In the field of anti money laundering(AML)in China,the people's Bank of China will regularly inspect the transaction data of financial companies and require them to upload the money laundering transaction data.The people's Bank of China compares the real money laundering data with the data uploaded by the company.If the company fails to report data or misreports,it will be fined a large amount.Among these data,stock trading data belongs to structured data,which is stored in relational database.Due to the complexity of table structure and the various relationships between tables,it is difficult to locate suspicious data accurately.Therefore,financial companies need test data to validate their trading compliance monitoring software.At present,the test data is constructed manually by the anti money laundering commissioner according to the relevant rules,which is difficult to model,difficult to generalize,and insufficient coverage.Based on the anti money laundering regulatory background of the people's Bank of China and the real needs of relevant financial companies,the thesis designs and implements AML rule based structural test data generation system,in order to replace the manual construction of test data mode.The solution of this technique is divided into four aspects.First,using natural language processing technique to extract the key parameters of user-defined rule text for automatic modeling.Second,providing users with the function of key parameter selection without refactoring code to improve the generalization of data generation.Third,using parametric combination rule technique to expand the algorithm parameter by parameter.And Fourth,analyzing the results of generated data to ensure the quality of test data.This system is divided into parametric combination rule module,data generation module,result analysis module and data management module.The server of the system is designed based on Django REST framework,which supports user-defined rule text and combination parameters.In the aspect of data generation,this thesis uses the cell and redis message queue to schedule the data generation task asynchronously.Finally,the generated data is stored in My SQL database for display and operation of the result analysis module and data management module.Experimental cases show that the technique can generate high-quality and comprehensive test data in a reasonable time,and can comprehensively test the transaction compliance monitoring software of financial companies,and optimize the accuracy of the software for money laundering data detection.