Application Of Random Forest And BP Neural Network In Anti-money Laundering Supervision

With the development of the economy,the increasing complexity of the financial system and the rapid development of Internet technology,there are increasingly signs of rampant domestic and international money laundering activities.Money laundering refers to the process of turning illegal income into a legally beneficial process through a series of financial trading operations.From the characteristics of money laundering transactions,these transactions have certain pre-planning,concealment and specificity.Through careful planning,criminals conduct a large number of superficial legal transactions to cover up the reality of their illegality.At present,all countries in the world attach great importance to anti-money laundering.The supervision of anti-money laundering has received wide attention from all sectors of society.The People's Bank of China has introduced various regulatory policies against anti-money laundering to identify anti-money laundering transactions.In accordance with the regulations of the People's Bank of China,various banks have carried out screening and monitoring of anti-money laundering according to law.However,at present,the anti-money laundering supervision of commercial banks mainly relies on a large number of manual operations,and has the characteristics of large workload,cumbersome steps,low efficiency,and prone to misjudgment.With the development of artificial intelligence and data mining technology,the use of advanced technology to automatically identify money laundering transactions has become a work that various banking institutions have paid more and more attention to.This paper takes LS Commercial Bank as the research object,first analyzes the status quo and problems of the bank's anti-money laundering supervision.Secondly,in order to establish a data mining model to identify money laundering transactions,this paper combines relevant literature,theoretical research and practical work design.The anti-money laundering transaction supervision index system;thirdly,this paper establishes a random forest and BP neural network model for empirical analysis,and compares it with the current manual preliminary screening process;finally,it puts forward relevant suggestions on the empirical analysis results.The conclusions of this paper are as follows:(1)At present,the anti-money laundering process of LS Commercial Bank has backward information technology,relying on a large number of manual rules to judge,and the recognition rate is low;(2)In order to solve the LS commercial bank's anti-money laundering,there are few regulatory indicators.In this paper,with reference to relevant literature,theory and practical experience,this paper designs twenty-three indicators of three dimensions: customer information,customer behavior,and transaction characteristics.These indicators are used as feature inputs of data mining models to characterize money laundering activities.The random forest model ranks the importance of these indicators,verifies the validity of the design indicators in this paper,and enriches the current regulatory indicator system.(3)In order to solve the current LS commercial banks,the anti-money laundering relies heavily on manual rules and the recognition rate is low.The problem is that this paper establishes a random forest model and a BP neural network model to identify anti-money laundering transactions.The performances of random forest model,BP neural network model and artificial preliminary screening were compared by empirical analysis.The random forest was stable and the prediction accuracy was91.7%.The prediction accuracy of BP neural network was 84.6%.The initial screening was accurate.The rate is 75%.From the perspective of operational efficiency,artificial initial screening takes much longer than using the random forest model and the BP neural network model.Therefore,through the empirical analysis in this chapter,we can see that the random forest and BP neural network models have a certain improvement in the accuracy of prediction compared with the manual preliminary screening,while the random forest and BP neural network are in the test set in terms of operational efficiency.The running time is less than0.05 seconds,and the initial judgment cost is 7210 seconds,and the recognition efficiency has been greatly improved.(4)Based on the empirical modeling analysis,this paper proposes that the bank can optimize the internal supervision process of anti-money laundering and apply it more.The anti-money laundering technology and the improvement of the anti-money laundering supervision index system have further optimized the current anti-money laundering supervision of LS Commercial Bank.Through the research of this paper,it provides an important reference for the introduction of advanced technologies such as data mining and artificial intelligence into the automatic prediction of anti-money laundering.