Application Research Of Anti-money Laundering System In Insurance Companies Based On Decision Tree Analysis

With the prosperous development of the financial industry, China’s financial sectoris facing more and more security challenge. Various forms of illegal income are gettinginto the financial system. The rapid growth of money laundering crimes brings greatharm to the country and society. It becomes increasingly important to do theanti-money laundering activity monitoring. Criminals make illegal income into banks,insurance companies and other financial institutions, and illegal income becomes legal.At present, the monitoring of anti-money laundering in China’s insuranceindustry stays in selection criteria set by Chinese law and regulations requirements.However, insurance companies have no innovation in filtrating large-value andsuspicious transactions. This paper focuses on the application of the anti-moneylaundering monitoring research in China’s insurance industry. Based on some maindate mining technologies, this paper focuses on the decision tree methods accordingto the established law and empirical data of anti-money laundering from insurancecompanies. The main works of the paper are as follows:Three algorithms of decision tree approach in data mining, which are CHAID,CART and QUEST, are used to identify the factors that influence the result oflarge-value and suspicious transactions. And then, this paper optimizes the model bysetting up the error cost to get better goodness of model fitting. The results of thethree algorithms on the fit of the model are basically same, but the accuracy of thelarge-value and suspicious transactions failed to reach ideal level. Therefore, thisarticle optimizes the model by setting up the error cost to makes the fit of the modelbetter. Eventually, the error rate of insurance companies treating the large-value andsuspicious transactions as good credit ones becomes less than15%.The Intervals between insured and surrender and Prem_mode are added in thevariables of large-value and suspicious transactions and applied to the anti-moneylaundering monitoring model. By analyzing the difference in the filtering results ofthe two systems, it proves that the improved system could enhance the coverage andaccuracy of the filtering. At last, this article provides suggestions from aspects ofpolicy and technology of anti-money laundering in insurance industry and showpracticability to some extent.