基于多层神经网络的洗钱风险评估方法

Money Laundry Risk Evaluation Method Based on Multi-level Neural Network

为了侦破采用信息技术手段的犯罪活动,需要强大的计算机智能系统。为此,提出一种利用神经网络,对银行客户潜在洗钱风险进行分类的方法,作为完整系统的部分支持。利用主元分析确定最合适的数据集,依靠L-M和贝叶斯正则化方法来训练最优效果的网络。实验结果表明,神经网络在解决目标问题的过程中比较有效。

Computer intelligent system is needed to crack crime activities using information technologies. This paper proposes a study aiming at constructing an effective anti money laundering system together with other respectable researches. A precise mode of BP network is constructed to evaluate the potential risk of money laundering of a certain bank account. Principle components analysis gives an inside view of data structure helping to find better input form for network. Levenberg-Marquardt algorithm accelerates the training process of BP impressively. And on the way generalization Bayesian regularization proves its value. Experimental result of the final system is satisfactory.