Research And Application Of AML Model Based On Bank Transaction Data

Money laundering is often associated with criminal activities such as drug trafficking.The characteristics of money laundering are different from normal trading behaviors,which is the basis of anti-money laundering work.With the development of economy,money laundering methods have been continuously upgraded,and data mining technology is also constantly developing.The purpose of this thesis is to use data mining technology in anti-money laundering work.Firstly,the thesis introduces the concept of money laundering and analyzes the current research status of data mining technology in the field of anti-money laundering,and summarizes the theoretical basis for building models.A two-stage clustering model based on hierarchical clustering and its improvement is the core of this thesis.By using two complementary clustering algorithms to build a two-stage clustering model,the model can not only handle outliers,but also effectively identify money laundering groups.The money laundering features are divided into transaction features and structural features.The transaction features can be obtained from the original transaction records,and the spatial features are difficult to obtain directly.The graph convolution neural network is used to extract the structural features of the nodes in the bank transaction network graph,and the feature fusion technology is used to fuse the transaction features and structural features as the basis for identifying money laundering data.The model is implemented by python3.6 in the platform named PyCharm which installed in win10 and is applied to a real transaction data set.The experimental results show that two-stage clustering model can effectively identify suspicious transactions,and the suspicious transaction recognition method using fusion features performs better in detection rate and false negative rate,the model proposed in this thesis can improve the efficiency of suspicious transaction identification.Finally,this thesis applies the model to the existing financial data analysis platform,reducing the total number of identifications by 321,the false alarm rate by 10.67% and 15% reduction in false negatives,effectively improving the efficiency of anti-money laundering staff.