The Anti-money Laundering Monitoring Research Based On Data Mining Technology

In today’s international society, money laundering risk becomes more and moreinternationalized. Data mining technology is one of the most effective means of identificationtechnology of anti-money laundering work. How to use data mining technology to improvethe efficiency of anti-money laundering information monitoring is a hot topic at present.In-depth study of recognition, rules and characteristics of money laundering of moneylaundering tracking, monitoring, blocking all has the important practical significance andapplication value and prevention.At present, commercial banks lack of money laundering monitoring technology,microcosmic mechanism this paper money laundering behavior from the research, exploreeffective technique method to identify money laundering. And the main choice ofmathematical statistics, association analysis, decision tree three algorithms, realize themonitoring and analysis of money laundering behavior data mining perspective.Through the transaction amount filtering money laundering, this paper constructs themonitoring model based on mathematical statistics. The main content in the hypotheticalpopulation a distribution, parameters estimation value to it, such as the variance. And theestimate results satisfy a certain degree of confidence. Algorithm according to the parametervalues whether the definition of transaction amount should be filtered. The simulationexperiment of money laundering in the mining of association rules, the improved AprioriTidalgorithm can type association on the trading behavior of abstraction model, at the same timeto complete the screening of large items in the user specified minimum support and minimumconfidence under. Finally according to the algorithm to calculate the correlation betweenmoney laundering behavior. In the simulation experiment of transaction classificationdecision tree, this paper combines the ID3algorithm to calculate the amount of informationdata set properties of two binary decision tree, and calculate the maximum gain according tothe amount of information, so that the node selection basis. Non class attribute that has thelargest gain should be used as the current node. In this paper, the realization of the above threealgorithms, and the results are analyzed.analysis.